Reduced false positive identification for spectroscopic quantification

Abstract

The device may receive information identifying the results of spectroscopic measurements performed on an unknown sample. The device may determine the judgment boundary of the quantification model based on configurable parameters such that a first plurality of training set samples of the quantification model are within the judgment boundary and a second plurality of training set samples of the quantification model are not within the judgment boundary. The device may determine the distance metric of the spectroscopic measurement performed on the unknown sample with respect to the judgment boundary. The device may determine a plurality of distance metrics of the second plurality of training set samples of the quantification model with respect to the judgment boundary. The device may provide information indicating whether the spectroscopic measurement performed on the unknown sample corresponds to the quantification model.

Description

Technology Field

[0001] The present invention relates to reduced false positive identification for spectroscopic quantification. Background Technology

[0002] Identification of raw materials can be utilized for the quality control of pharmaceuticals. For example, the identification of raw materials of a medical substance may be performed to determine whether the components of the medical substance correspond to the packaging label associated with the medical substance. Similarly, the quantification of raw materials may be performed to determine the concentration of a specific component in a specific sample. For example, the quantification of raw materials may be performed to determine the concentration of an active ingredient within a pharmaceutical. Spectroscopy can enable non-destructive identification and / or quantification of raw materials with reduced preparation and data acquisition times compared to other chemical analysis techniques.

[0003] According to a first possible embodiment, the device may include one or more memories communicably coupled to one or more processors. The one or more memories and the one or more processors may be configured to receive information identifying the results of spectroscopic measurements performed on an unknown sample. The one or more memories and the one or more processors may be configured to determine the decision boundary of the quantification model based on configurable parameters such that a first plurality of training set samples of the quantification model are within the decision boundary and a second plurality of training set samples of the quantification model are not within the decision boundary. The one or more memories and the one or more processors may be configured to determine a distance metric of the spectroscopic measurements performed on the unknown sample with respect to the decision boundary. The one or more memories and the one or more processors may be configured to determine a plurality of distance metrics of the second plurality of training set samples of the quantification model with respect to the decision boundary. The one or more memories and the one or more processors may be configured to determine whether a spectroscopic measurement performed on the unknown sample corresponds to the quantification model based on the distance metric of the spectroscopic measurement and the multiple distance metrics of the second plurality of training set samples. The one or more memories and the one or more processors may be configured to provide information indicating whether a spectroscopic measurement performed on the unknown sample corresponds to the quantification model.

[0004] More specifically, in additional embodiments, The above one or more processors are configured to determine that the spectroscopic measurement performed on the unknown sample does not correspond to the quantification model when determining whether the spectroscopic measurement performed on the unknown sample corresponds to the quantification model; and the above one or more processors are configured to provide information indicating that the spectroscopic measurement does not correspond to the quantification model when providing information indicating whether the spectroscopic measurement performed on the unknown sample corresponds to the quantification model. Additionally, the above one or more processors are configured to determine that the spectroscopic measurement corresponds to the quantification model when determining whether the spectroscopic measurement performed on the unknown sample corresponds to the quantification model; and the above one or more processors are configured to provide information indicating that the spectroscopic measurement corresponds to the quantification model when providing information indicating whether the spectroscopic measurement performed on the unknown sample corresponds to the quantification model. Additionally, the above one or more processors determine a statistical metric of the distance metric for the plurality of distance metrics when determining whether the spectroscopic measurement performed on the unknown sample corresponds to the quantification model; Based on the above statistical metric, it is configured to determine whether a spectroscopic measurement performed on the above unknown sample corresponds to the above quantification model. Additionally, the statistical metric is a log-normal standard deviation; and the one or more processors are configured to determine that the log-normal standard deviation satisfies a threshold value when determining whether a spectroscopic measurement performed on the above unknown sample corresponds to the above quantification model based on the above statistical metric; and to determine whether a spectroscopic measurement performed on the above unknown sample corresponds to the above quantification model based on the determination that the log-normal standard deviation satisfies the threshold value. Additionally, the above quantification model is associated with a single class support vector machine (SC-SVM) classifier.Additionally, the one or more processors are further configured to receive a plurality of spectroscopic measurements associated with the first plurality of training set samples and the second plurality of training set samples; determine the quantification model based on the plurality of spectroscopic measurements; verify the quantification model based on another plurality of spectroscopic measurements of the plurality of verification set samples; and store the quantification model, and the one or more processors are further configured to obtain the quantification model from the storage medium when determining the judgment boundary; and determine the judgment boundary after obtaining the quantification model from the storage medium.

[0005] According to some possible second embodiment, a non-transient computer-readable medium may store one or more commands. When executed by one or more processors, the one or more commands may cause the acquisition of a quantification model associated with a specific type of substance of interest. The quantification model may be configured to determine the concentration of a specific component in a sample of the specific type of substance of interest. When executed by one or more processors, the one or more commands may cause the one or more processors to receive information identifying the results of a specific spectroscopic measurement performed on an unknown sample. When executed by one or more processors, the one or more commands may cause the one or more processors to aggregate other spectroscopic measurements of the training set samples of the quantification model into a single class of the quantification model. When executed by one or more processors, the one or more commands may cause the one or more processors to subdivide the other spectroscopic measurements of the training set samples into a first group and a second group. The first group of the other spectroscopic measurements may be within a determination boundary. The second group of the other spectroscopic measurements may not be within the determination boundary. When executed by the one or more processors, the one or more commands may cause the one or more processors to determine that the metric of a specific spectroscopic measurement performed on the unknown sample satisfies a threshold value for the corresponding metric of the second group of the other spectroscopic measurements. When executed by the one or more processors, the one or more commands may cause the one or more processors to provide information indicating that the unknown sample is not the specific type of substance of interest.

[0006] More specifically, in additional embodiments, the unknown sample is a substance of a different type from the specific type of substance of interest. Also, the unknown sample is the specific type of substance of interest and is an inaccurately obtained measurement. The metric and the corresponding metric are judgment values. Also, the threshold is a threshold amount of the standard deviation of the metric from the mean of the corresponding metric. Also, the metric and the corresponding metric are determined using a single-class support vector machine technique. Also, the quantification model is a local model, and the one or more instructions, when executed by the one or more processors, cause the one or more processors to further perform the operation of performing a first determination related to a specific spectroscopic measurement of the unknown sample using a global model related to the specific type of substance of interest; and the operation of generating the local model using the in-situ local modeling technique based on a specific result of the first determination; and the one or more instructions causing the one or more processors to obtain the quantification model cause the one or more processors to obtain the quantification model based on generating the local model.

[0007] According to some possible third embodiment, the method may include the step of receiving by the device information identifying the result of a near-infrared (NIR) spectroscopic measurement performed on an unknown sample. The method may include the step of determining a determination boundary of a quantification model by the device, wherein the determination boundary divides a single class of the quantification model into a first plurality of training set samples of the quantification model within the determination boundary and a second plurality of training set samples of the quantification model not within the determination boundary. The method may include the step of determining by the device that a specific distance metric of the NIR spectroscopic measurement performed on the unknown sample satisfies a threshold value for another distance metric of the second plurality of training set samples. The method may include the step of providing by the device information indicating that the NIR spectroscopic measurement performed on the unknown sample does not correspond to the quantification model, based on the determination that the specific distance metric of the NIR spectroscopic measurement performed on the unknown sample satisfies a threshold value for another distance metric of the second plurality of training set samples.

[0008] More specifically, in additional embodiments, the method further comprises the steps of: determining the type of the unknown sample based on the NIR spectroscopic measurements using a classification model and determining that a specific distance metric performed on the unknown sample satisfies the threshold for another distance metric of the second plurality of training set samples; and providing information for identifying the type of the unknown sample. The method further comprises the step of determining the determination boundary based on a kernel function. The kernel function is at least one of a radial basis function, a polynomial function, a linear function, or an exponential function. The threshold exceeds at least one of one standard deviation, two standard deviations, or three standard deviations. The first plurality of training set samples and the second plurality of training set samples are associated with a set of component concentrations, and each concentration of the component in the set of component concentrations is associated with a threshold amount of the training set samples of the first plurality of training set samples and the second plurality of training set samples. Brief explanation of the drawing

[0009] FIGS. 1a and 1b are schematic diagrams of exemplary embodiments described herein; FIG. 2 is a drawing illustrating an exemplary environment in which the system and / or method described in this specification may be implemented; FIG. 3 is a drawing showing exemplary components of one or more devices of FIG. 2; FIG. 4 is a flowchart of an exemplary process for generating a quantification model for spectroscopic quantification; FIG. 5 is a drawing illustrating an exemplary embodiment of the exemplary process illustrated in FIG. 4; FIG. 6 is a flowchart of an exemplary process for avoiding false positive identification during spectroscopic quantification; and FIGS. 7A and FIGS. 7B are drawings illustrating exemplary embodiments of the exemplary process shown in FIG. 6. Specific details for implementing the invention

[0010] The following detailed description of exemplary embodiments refers to the attached drawings. In different drawings, the same reference numerals may represent the same or similar elements.

[0011] Raw material identification (RMID) is a technique used to identify the components (e.g., constituents) of a specific sample for identification, verification, etc. For example, RMID can be used to verify whether the constituents of a pharmaceutical substance correspond to the set of constituents identified on the label. Similarly, raw material quantification is a technique used to perform quantitative analysis on a specific sample, for example, to determine the concentration of a specific component substance in a specific sample. A spectrometer can be used to perform spectroscopic analysis on a sample (e.g., a pharmaceutical substance) to determine the components of the sample, the concentration of the components of the sample, etc. The spectrometer can determine a set of measurements of the sample and can provide a set of measurements for spectroscopic determination. Spectroscopic classification techniques (e.g., a classifier) ​​can make it possible to determine the components of the sample or the concentration of the components of the sample based on the set of measurements of the sample.

[0012] However, some unknown samples undergoing spectroscopic quantification do not actually fall within the class of substances for which the quantification model is configured to be quantified. For example, in the case of a quantification model trained to determine the concentration of a specific type of protein in a fish sample, the user may unintentionally provide a beef sample for quantification. In this case, the control unit can perform spectroscopic quantification of the beef sample and identify the beef sample as having a specific concentration of a specific type of protein. However, due to differences in the spectral signatures between beef and fish, and between their proteins, the identification may be inaccurate, which can be referred to as false positive identification.

[0013] As another example, a quantification model can be trained to quantify the relative concentrations of different types of sugars (e.g., glucose, lactose, galactose, etc.) and unknown samples. However, the user of the spectrometer and control unit may unintentionally attempt to classify an unknown sample of sugars based on the inaccurate use of the spectrometer to perform measurements. For example, the user may operate the spectrometer with an inaccurate distance from the unknown sample, under environmental conditions different from the calibration conditions performed to train the quantification model, and / or under other conditions that result in inaccurately obtained measurements. In this case, the control unit may receive an inaccurate spectrum for the unknown sample, which may lead to a false positive identification of the unknown sample as having a first type of sugar at a first concentration when the unknown sample actually contains a second type of sugar at a second concentration.

[0014] Some embodiments described herein may use Single Class Support Vector Machine (SC-SVM) technology to reduce the possibility of false positive identification during spectroscopic quantification. For example, a control device receiving spectroscopic measurements of an unknown sample may determine whether the spectroscopic measurements of the unknown sample correspond to a class of substance configured to be quantified by the spectroscopic model. In some embodiments, the control device may determine that the unknown sample is not associated with the class of substance configured to be quantified by the spectroscopic model, and may avoid false positive identification of the unknown sample by providing information indicating that the unknown sample is not associated with the class of substance. Alternatively, based on the determination that the unknown sample is associated with the class of substance configured to be quantified by the spectroscopic model, the control device may analyze the spectrum of the unknown sample to provide spectroscopic determinations, for example, concentration, classification, etc. Additionally, the control device may use confidence metrics, such as probability estimation and judgment values, to filter out false positive identification.

[0015] In this way, the accuracy of the spectral analysis is improved compared to spectral analysis performed without identifying potential error samples (e.g., samples associated with a class of substance for which a spectral model is constructed, or samples for which spectroscopic measurements were obtained inaccurately) and / or reliability metrics. Furthermore, the determination of whether a substance is associated with a class for which a spectral model is constructed can be used when generating a quantification model based on a training set of known spectral samples. For example, the control unit may determine that a sample in the training set is not of a type corresponding to the remaining samples in the training set (e.g., based on human error that causes inaccurate samples to be introduced into the training set) and may decide not to include data regarding that sample when generating the quantification model. In this manner, the control unit improves the accuracy of the quantification model for spectral analysis.

[0016] FIGS. 1a and FIGS. 1b are schematic diagrams of an exemplary embodiment (100) described herein. As shown in FIG. 1a, the exemplary embodiment (100) may include a control device and a spectrometer.

[0017] As further illustrated in FIG. 1a, the control device may cause the spectrometer to perform a series of spectroscopic measurements on a training set and a validation set (e.g., a known set of samples used for training and validating a classification model). The training set and the validation set may be selected to contain a threshold amount of samples for the component to be trained on. The substance that may occur and can be used to train the quantification model may be referred to as the substance of interest. In this case, the training set and the validation set may include, for example, a first group of samples representing a first concentration of the substance of interest, a second group of samples representing a second concentration of the substance of interest, etc., to enable the training of the quantification model to identify the concentration of the substance of interest in unknown samples.

[0018] As further illustrated in FIG. 1a, the spectrometer can perform a series of spectroscopic measurements on the training set and the validation set based on commands received from the control unit. For example, the spectrometer can determine the spectrum for each sample in the training set and the validation set so that the control unit can generate a set of classes to classify an unknown sample as one of the substances of interest in the quantification model.

[0019] The spectrometer can provide a set of spectroscopic measurements to the control unit. The control unit can generate a quantification model based on the set of spectroscopic measurements using a specific determination technique. For example, the control unit can generate a quantification model using a Support Vector Machine (SVM) technique, such as a Single Class SVM (SC-SVM) technique (e.g., machine learning techniques for information determination). The quantification model may include information associated with assigning a specific spectrum to a specific concentration of a component of the substance of interest (e.g., a specific concentration level of the component in the substance of interest). In this way, the control unit can provide information for identifying the concentration of a component in an unknown sample based on assigning the spectrum of the unknown sample to a specific concentration class of the quantification model corresponding to the specific concentration.

[0020] As illustrated in FIG. 1b, the control device may receive a quantification model (e.g., from a storage component, from another control device that generated the quantification model, etc.). The control device may have a spectrometer perform a series of spectroscopic measurements on an unknown sample (e.g., an unknown sample to be classified or quantified). The spectrometer may perform a series of spectroscopic measurements based on commands received from the control device. For example, the spectrometer may determine the spectrum for the unknown sample. The spectrometer may provide a set of spectroscopic measurements to the control device. The control device may attempt to quantify the unknown sample based on the quantification model (e.g., classifying the unknown sample into a specific class associated with a specific concentration or specific amount of a specific component in the unknown sample). For example, the control device may attempt to determine a specific concentration of ibuprofen in an unknown sample (e.g., a pill), a specific unit amount of glucose in an unknown sample (e.g., a sugar-based product), etc.

[0021] With respect to FIG. 1b, the control device may attempt to determine whether an unknown sample corresponds to a quantification model. For example, the control device may determine a confidence metric corresponding to the probability that the unknown sample belongs to a substance of interest (e.g., any concentration among the concentration sets constructed by the quantification model using a training set and a validation set). As an example, in the case of a quantification model configured to identify the concentration of ibuprofen in a sample of ibuprofen pills, the control device may determine whether the unknown sample is an ibuprofen pill (and not other types of pills such as acetaminophen pills, acetylsalicylic acid pills, etc.). As another example, in the case of a quantification model configured to identify the concentration of salt in fish meat, the control device may determine whether the unknown sample is fish meat (and not chicken, beef, pork, etc.).

[0022] In this case, based on the control device determining that a reliability metric, such as a probability estimation or the output of a support vector machine's decision value, satisfies a threshold value (e.g., the standard deviation threshold described herein), the control device may determine that the unknown sample is not the substance of interest (e.g., that the unknown sample is a different substance, or that the spectroscopic measurement of the unknown sample was performed inaccurately). In this case, the control device may reduce the possibility of the unknown sample being misidentified as belonging to a specific concentration of a component of the substance of interest by reporting that the unknown sample cannot be accurately quantified using a quantification model.

[0023] In this way, the control device enables spectroscopy of an unknown sample with enhanced accuracy compared to other quantification models, based on reducing the possibility that an unknown sample may be falsely identified and reported as a specific concentration of a component in the substance of interest.

[0024] As mentioned above, FIGS. 1a and 1b are provided merely as examples. Other examples are possible and may differ from those described in relation to FIGS. 1a and 1b.

[0025] FIG. 2 is a drawing illustrating an exemplary environment (200) in which the system and / or method described herein may be implemented. As illustrated in FIG. 2, the environment (200) may include a control device (210), a spectrometer (220), and a network (230). Devices in the environment (200) may be interconnected via a wired connection, a wireless connection, or a combination of a wired connection and a wireless connection.

[0026] The control device (210) includes one or more devices capable of storing, processing, and / or routing information related to spectroscopic quantification. For example, the control device (210) may include a server, computer, wearable device, cloud computing device, etc., which uses the quantification model to generate a quantification model based on a set of measurements in a training set, validate the quantification model based on a set of measurements in a validation set, and / or perform spectroscopic quantification based on a set of measurements of an unknown sample. In some embodiments, the control device (210) may use machine learning techniques to determine whether spectroscopic measurements of an unknown sample should be classified as not corresponding to the substance of interest of the quantification model as described herein. In some embodiments, the control device (210) may be associated with a specific spectrometer (220). In some embodiments, the control device (210) may be associated with a plurality of spectrometers (220). In some embodiments, the control device (210) may receive information from another device in the environment (200), such as a spectrometer (220), and / or transmit information to another device.

[0027] The spectrometer (220) includes one or more devices capable of performing spectroscopic measurements on a sample. For example, the spectrometer (220) may include a spectrometer device that performs spectroscopic analysis (vibrational spectroscopic analysis, e.g., near-infrared (NIR) spectroscopy, mid-infrared (mid-IR) spectroscopy, Raman spectroscopy, etc.). In some embodiments, the spectrometer (220) may be integrated into a wearable device, such as a wearable spectrometer. In some embodiments, the spectrometer (220) may receive information from and / or transmit information to another device within the environment (200), such as a control device (210).

[0028] The network (230) may include one or more wired networks and / or wireless networks. For example, the network (230) may include a cellular network (e.g., LTE (long-term evolution) network, 3G network, Code Division Multiple Access (CDMA) network, etc.), a public land mobile network (PLMN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a telephone network (e.g., a public switched telephone network (PSTN)), a private network, an ad hoc network, an intranet, the internet, a fiber-optic-based network, a cloud computing network, etc., and / or a combination of these or other types of networks.

[0029] The number and arrangement of devices and networks shown in FIG. 2 are provided merely as an example. In practice, there may be additional devices and / or networks, fewer devices and / or networks, different devices and / or networks, or devices and / or networks arranged differently than those shown in FIG. 2. Additionally, two or more devices shown in FIG. 2 may be implemented within a single device, or the single device shown in FIG. 2 may be implemented as a number of distributed devices. For example, although the control device (210) and the spectrometer (220) are described herein as two separate devices, the control device (210) and the spectrometer (220) may be implemented within a single device. Additionally or alternatively, a set of devices (e.g., one or more devices) of the environment (200) may perform one or more functions described as being performed by another set of devices of the environment (200).

[0030] FIG. 3 is a drawing illustrating exemplary components of a device (300). The device (300) may correspond to a control device (210) and / or a spectrometer (220). In some embodiments, the control device (210) and / or the spectrometer (220) may include one or more devices (300) and / or one or more components of the device (300). As illustrated in FIG. 3, the device (300) may include a bus (310), a processor (320), a memory (330), a storage component (340), an input component (350), an output component (360), and a communication interface (370).

[0031] The bus (310) includes components that allow communication between components of the device (300). The processor (320) is implemented in hardware, firmware, or a combination of hardware and software. The processor (320) is a central processing unit (CPU), a graphics processing unit (GPU), an acceleration processing unit (APU), a microprocessor, a microcontroller, a digital signal processor (DSP), a field programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or other types of processing components. In some embodiments, the processor (320) includes one or more processors that can be programmed to perform functions. Memory (330) includes random access memory (RAM), read-only memory (ROM), and / or other types of dynamic or static storage devices (e.g., flash memory, magnetic memory, and / or optical memory) that store information and / or instructions for use by the processor (320).

[0032] The storage component (340) stores information and / or software related to the operation and use of the device (300). For example, the storage component (340) may include a hard disk (e.g., magnetic disk, optical disk, magneto-optical disk and / or solid-state disk), a compact disk (CD), a digital versatile disc (DVD), a floppy disk, a cartridge, a magnetic tape, and / or other types of non-transient computer-readable media together with a corresponding drive.

[0033] The input component (350) includes a component that allows the device (300) to receive information, for example, through user input (e.g., a touch screen display, keyboard, keypad, mouse, button, switch, and / or microphone). Additionally or alternatively, the input component (350) may include a sensor for detecting information (e.g., a global positioning system (GPS) component, accelerometer, gyroscope, and / or actuator). The output component (360) includes a component that provides output information from the device (300) (e.g., a display, speaker, and / or one or more light-emitting diodes (LEDs)).

[0034] The communication interface (370) includes a transceiver-type component (e.g., a transceiver and / or separate receiver and transmitter) that enables the device (300) to communicate with another device, for example, through a wired connection, a wireless connection, or a combination of a wired connection and a wireless connection. The communication interface (370) may enable the device (300) to receive information from another device and / or provide information to another device. For example, the communication interface (370) may include an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a Universal Serial Bus (USB) interface, a wireless short-range network interface, a cellular network interface, etc.

[0035] The device (300) may perform one or more processes described herein. The device (300) may perform these processes based on a processor (320) that executes software instructions stored by a non-transient computer-readable medium, such as a memory (330) and / or a storage component (340). A computer-readable medium is defined herein as a non-transient memory device. A memory device may include a memory space within a single physical storage device or a memory space distributed across multiple physical storage devices.

[0036] Software instructions may be read into memory (330) and / or storage components (340) from another computer-readable medium or from another device via a communication interface (370). When executed, software instructions stored in memory (330) and / or storage components (340) may cause the processor (320) to perform one or more processes described herein. Additionally or alternatively, hardwired circuitry may be used instead of or together with software instructions to perform one or more processes described herein. Accordingly, the embodiments described herein are not limited to any specific combination of hardware circuitry and software.

[0037] The number and arrangement of components shown in FIG. 3 are provided as an example. In practice, the device (300) may include additional components, a smaller number of components, different components, or components arranged differently compared to those shown in FIG. 3. Additionally or alternatively, a set of components of the device (300) (e.g., one or more components) may perform one or more functions described as being performed by another set of components of the device (300).

[0038] FIG. 4 is a flowchart of an exemplary process (400) for generating a quantification model for spectroscopic quantification. In some embodiments, one or more process blocks of FIG. 4 may be performed by a control device (210). In some embodiments, one or more process blocks of FIG. 4 may be performed by a device or group of devices different from the control device (210), such as a spectrometer (220), or including this control device.

[0039] As illustrated in FIG. 4, the process (400) may include the step of performing a series of spectroscopic measurements on a training set and / or a validation set (Block 410). For example, a control unit (210) may cause a spectrometer (220) to perform a series of spectroscopic measurements on the training set and / or validation set of samples to determine the spectrum for each sample of the training set and / or validation set (e.g., using a processor (320), a communication interface (370), etc.). The training set may represent a set of samples of one or more known substances having a set of concentrations of a component, which is used to generate a quantification model for the component. Similarly, the validation set may represent a set of samples of one or more known substances having a set of concentrations of a component, which is used to verify the accuracy of the quantification model. For example, the training set and / or validation set may include one or more versions of a specific substance with different sets of concentrations (e.g., one or more versions manufactured by different manufacturers to control for manufacturing differences).

[0040] In some embodiments, the training set and / or validation set may be selected based on the expected set of substances of interest on which spectroscopic quantification is to be performed using a quantification model. For example, if spectroscopic quantification is expected to be performed on a pharmaceutical substance to determine the concentration of a specific component of the pharmaceutical substance, the training set and / or validation set may include a set of samples of the specific component with different possible concentrations in the set of pharmaceutical substances to be tested for the presence of the specific component.

[0041] In some embodiments, the training set and / or verification set may be selected to include a specific amount of samples for each concentration of the substance. For example, the training set and / or verification set may be selected to include a number of samples of a specific concentration (e.g., 5 samples, 10 samples, 15 samples, 50 samples, etc.). In this way, the control device (210) includes a threshold amount of the spectrum associated with a specific type of substance, thereby enabling the generation and / or verification of a class of quantification models (e.g., a group of samples corresponding to a specific concentration of a component) to which an unknown sample can be accurately assigned (e.g., based on an unknown sample having a specific concentration of a component).

[0042] In some embodiments, the control device (210) may cause a plurality of spectrometers (220) to perform a series of spectroscopic measurements to account for one or more physical conditions. For example, the control device (210) may cause a first spectrometer (220) and a second spectrometer (220) to perform a series of oscillatory spectroscopic measurements using NIR spectroscopy. Additionally or alternatively, the control device (210) may cause a series of spectroscopic measurements to be performed at multiple times, at multiple locations, under multiple different experimental conditions, etc. In this way, the control device (210) reduces the possibility that the spectroscopic measurements will become inaccurate as a result of physical conditions in relation to causing a series of spectroscopic measurements to be performed by a single spectrometer (220).

[0043] In this way, the control device (210) causes a series of spectroscopic measurements to be performed on the training set and / or verification set.

[0044] As further illustrated in FIG. 4, the process (400) may include the step of receiving information identifying the results of a set of spectroscopic measurements (Block 420). For example, a control device (210) may receive information identifying the results of a set of spectroscopic measurements (e.g., using a processor (320), a communication interface (370), etc.). In some embodiments, the control device (210) may receive information identifying a set of spectra corresponding to samples of a training set and / or a verification set. For example, the control device (210) may receive information identifying a specific spectrum observed when the spectrometer (220) performs spectral analysis on the training set. In some embodiments, the control device (210) may simultaneously receive information identifying the spectra of the training set samples and the verification set samples. In some embodiments, the control device (210) may receive information identifying the spectrum of a training set sample, may generate a quantification model, and may receive information identifying the spectrum of a verification set sample after generating the quantification model to enable testing of the quantification model.

[0045] In some embodiments, the control device (210) may receive information identifying the results of a set of spectroscopic measurements from a plurality of spectrometers (220). For example, the control device (210) may control physical conditions such as differences between the plurality of spectrometers (220), possible differences in experimental conditions, and / or the like by receiving spectroscopic measurements performed by the plurality of spectrometers (220), performed at a plurality of different times, performed at a plurality of different locations, and / or the like.

[0046] In some embodiments, the control device (210) may remove one or more spectra from those used when generating the quantification model. For example, the control device (210) may perform spectroscopic quantification and determine that the spectra do not correspond to the type of substance for which the quantification model is configured to quantify, or determine that the sample corresponding to the spectra was unintentionally a substance of interest (e.g., based on human error when performing spectroscopic quantification correctly, errors in information identifying the spectra of the training set, etc.). In this case, the control device (210) may determine to remove the spectra from the training set. In this way, the control device (210) may improve the accuracy of the quantification model by reducing the likelihood that the quantification model will be generated using incorrect or inaccurate information regarding the training set or the validation set.

[0047] In this way, the control device (210) receives information identifying the result of the set of spectroscopic measurements.

[0048] As further illustrated in FIG. 4, the process (400) may include the step (block 430) of generating a quantification model based on information identifying the results of a set of spectroscopic measurements. For example, the control unit (210) may generate a quantification model associated with SVM classifier technology based on information identifying the results of a set of spectroscopic measurements (e.g., using a processor (320), memory (330), storage component (340), etc.).

[0049] SVM may refer to a supervised learning model that performs pattern recognition and uses confidence metrics for quantification. In some embodiments, the control unit (210) may use a specific type of kernel function to determine the similarity of two or more inputs (e.g., spectra) when generating a quantification model using SVM techniques. For example, the control unit (210) may use a kernel function of the type of a radial basis function (RBF) (e.g., referred to as SVM-rbf) which can be expressed as k(x,y) = exp(-||xy||^2) for spectra x and y; a kernel function of the type of a linear function which can be expressed as k(x,y) = <x·y> (e.g., referred to as SVM-linear, and e.g., referred to as hier-SVM-linear when used in a multi-stage decision technique); a kernel function of the type of a sigmoid function of a kernel function; a kernel function of the type of a polynomial function; Exponential kernel functions, etc., may be used. In some embodiments, the control device (210) may generate a quantification model using a single-class SVM (SC-SVM) classifier technique. For example, the control device (210) may aggregate multiple classes corresponding to multiple concentrations of components within the training set to generate a single class representing the quantification model. In this case, the control device (210) may use a confidence metric to determine the likelihood that an unknown sample is of the type configured for the quantification model to analyze, as described herein.

[0050] In some embodiments, the control unit (210) may use a specific type of confidence metric of an SVM, such as a probability value-based SVM (e.g., a decision based on determining the probability that a sample is a member of a class of a set of classes of possible concentrations), or a decision value-based SVM (e.g., a decision using a decision function to vote that a sample is a member of a class of a set of classes). For example, while using a quantification model with a decision value-based SVM, the control unit (210) may determine whether an unknown sample is located within the boundaries of a component class (e.g., a specific amount or concentration of a component of the unknown sample) based on plotting the spectrum of the unknown sample, and may assign the sample to a class based on whether the unknown sample is located within the boundaries of the component class. In this way, the control unit (210) may determine whether to assign the unknown spectrum to a specific class for quantification.

[0051] Some embodiments described herein are described as a specific set of machine learning techniques, but other techniques for determining information regarding unknown spectra, such as the classification of substances, are also possible.

[0052] In some embodiments, the control device (210) may select a specific classifier to be used to generate a quantification model from a series of quantification techniques. For example, the control device (210) may generate multiple quantification models corresponding to multiple classifiers and may test multiple quantification models by determining, for example, the transferability of each model (e.g., the degree of accuracy with which a quantification model generated based on spectroscopic measurements performed at a first spectrometer (220) is applied to spectroscopic measurements performed at a second spectrometer (220)), large-scale decision accuracy (e.g., the accuracy with which a quantification model can be used to simultaneously identify concentrations for a sample amount satisfying a threshold value), etc. In such cases, the control device (210) may select a classifier such as an SVM classifier (e.g., a Hier-SVM linear classifier, an SC-SVM classifier, and / or others) based on determining that the classifier is associated with superior transferability and / or large-scale decision accuracy compared to other classifiers.

[0053] In some embodiments, the control device (210) may generate a quantification model based on information identifying samples of a training set. For example, the control device (210) may identify a spectral class having a type or concentration of a substance using information identifying a type or concentration of a substance represented by samples of a training set. In some embodiments, the control device (210) may train the quantification model when generating the quantification model. For example, the control device (210) may allow the quantification model to be trained using a portion of a set of spectroscopic measurements (e.g., measurements associated with the training set). Additionally or alternatively, the control device (210) may perform an evaluation of the quantification model. For example, the control device (210) may validate the quantification model (e.g., for predicted intensity) using another portion of the set of spectroscopic measurements (e.g., a validation set).

[0054] In some embodiments, the control unit (210) may verify the quantification model using a multi-stage decision technique. For example, in the case of in-situ local modeling-based quantification, the control unit (210) may determine that the quantification model is accurate when used in association with one or more local quantification models. In this way, the control unit (210) ensures that the quantification model is generated with critical accuracy before providing the quantification model for use by, for example, by the control unit (210), by another control unit (210) associated with another spectrometer (220), and / or etc.

[0055] In some embodiments, the control device (210) may provide the quantification model to another control device (210) associated with another spectrometer (220) after generating the quantification model. For example, the first control device (210) may generate the quantification model and provide the quantification model to the second control device (210) for use. In this case, for quantification based on local modeling at the source, the second control device (210) may store the quantification model (e.g., a global quantification model) and may use the quantification model when generating one or more local quantification models at the source to determine the concentration of a component of a substance in one or more samples of an unknown set. Additionally or alternatively, the control device (210) may store a quantification model for use by the control device (210), such as when performing quantification or when generating one or more local quantification models (e.g., in the case of local modeling-based quantification at the source). In this way, the control device (210) provides a quantification model for use in spectroscopic quantification of an unknown sample.

[0056] In this way, the control device (210) generates a quantification model based on information identifying the results of a set of spectroscopic measurements.

[0057] FIG. 4 illustrates an exemplary block of process (400), but in some embodiments, process (400) may include additional blocks, a smaller number of blocks, different blocks, or blocks arranged differently compared to that illustrated in FIG. 4. Additionally or alternatively, two or more blocks of process (400) may be performed in parallel.

[0058] FIG. 5 is a drawing illustrating an exemplary embodiment (500) of the exemplary process (400) illustrated in FIG. 4. FIG. 5 illustrates an example of generating a quantification model.

[0059] As illustrated in FIG. 5 and reference number (505), a control unit (210-1) transmits information to a spectrometer (220-1) to command the spectrometer (220-1) to perform a series of spectroscopic measurements on a training set and a verification set (510). It is assumed that the training set and the verification set (510) include a first set of training samples (e.g., measurements of the training samples are used to train a quantification model) and a second set of verification samples (e.g., measurements of the verification samples are used to verify the accuracy of the quantification model). As illustrated in reference number (515), the spectrometer (220-1) performs a series of spectroscopic measurements based on the received command. As illustrated in reference number (520), the control unit (210-1) receives a first set of spectra of the training samples and a second set of spectra of the verification samples. In this case, the training sample and the validation sample may include samples of multiple concentrations of a component in the group of substances of interest for quantification. For example, the control device (210-1) may generate a global model (e.g., a global classification model or a quantification model) to identify the type of meat using a local-local modeling technique (to generate a local model such as a local classification model or a quantification model) and receive a spectrum regarding the quantification of the concentration of a specific protein in the type of meat. In this case, the control device (210-1) may be configured to generate multiple local quantification models (e.g., a first quantification model to determine the concentration of a specific protein in a first type of meat identified using local-local modeling, a second quantification model to determine the concentration of a specific protein in a second type of meat identified using local-local modeling, etc.). It is assumed that the control device (210-1) stores information identifying each sample of the training set and the validation set (510).

[0060] With reference to FIG. 5, it is assumed that the control unit (210-1) chooses to use a hier-SVM linear classifier to generate a classification model and an SC-SVM classifier for multiple quantification models. As illustrated by reference number (525), the control unit (210-1) trains a global classification model using a hier-SVM linear classifier and a first set of spectra, and verifies the global classification model using a hier-SVM linear classifier and a second set of spectra. Additionally, the control unit (210-1) trains and verifies multiple local quantification models (e.g., local quantification models corresponding to each class of the global classification model and / or each class of the local classification model generated based on the global classification model). The control unit (210-1) assumes that it determines that the quantification model satisfies a verification threshold (e.g., has an accuracy exceeding the verification threshold). As illustrated by reference number (530), the control unit (210-1) provides a quantification model to the control unit (210-2) (for use when performing quantification on spectroscopic measurements performed by the spectrometer (220-2), for example) and provides a quantification model to the control unit (210-3) (for use when performing quantification on spectroscopic measurements performed by the spectrometer (220-3).

[0061] As previously mentioned, FIG. 5 is provided merely as an example. Other examples are possible and may differ from those described in relation to FIG. 5.

[0062] In this way, the control device (210) enables the generation of a quantification model based on a selected classification technique (e.g., selected based on the model's exclusivity, large-scale quantification accuracy, etc.) and the distribution of the quantification model for use by one or more other control devices (210) associated with one or more spectrometers (220).

[0063] FIG. 6 is a flowchart of an exemplary process (600) for avoiding false positive identification while quantifying raw materials. In some embodiments, one or more process blocks of FIG. 6 may be performed by a control device (210). In some embodiments, one or more process blocks of FIG. 6 may be performed by a device or group of devices different from the control device (210), such as a spectrometer (220), or including this control device.

[0064] As illustrated in FIG. 6, the process (600) may include the step of receiving information identifying the results of a set of spectroscopic measurements performed on an unknown sample (Block 610). For example, a control device (210) may receive information identifying the results of a set of NIR spectroscopic measurements performed on an unknown sample (e.g., using a processor (320), a communication interface (370), etc.). In some embodiments, the control device (210) may receive information identifying the results of a set of spectroscopic measurements on an unknown set (e.g., a plurality of samples). The unknown set may include a set of samples (e.g., unknown samples) on which a determination (e.g., spectroscopic quantification) is to be performed. For example, the control device (210) may have a spectrometer (220) perform a series of spectroscopic measurements on a set of unknown samples and may receive information identifying a set of spectra corresponding to the set of unknown samples.

[0065] In some embodiments, the control device (210) may receive information identifying results from a plurality of spectrometers (220). For example, the control device (210) may have a plurality of spectrometers (220) perform a series of spectroscopic measurements on an unknown set (e.g., the same set of samples) and may receive information identifying a set of spectra corresponding to samples of the unknown set. Additionally or alternatively, the control device (210) may receive information identifying results of a set of spectroscopic measurements performed at a plurality of times, at a plurality of locations, and / or, etc., and may quantify a specific sample based on a set of spectroscopic measurements performed at a plurality of times, at a plurality of locations, and / or, etc. (e.g., based on averaging a set of spectroscopic measurements or based on other techniques). In this way, the control device (210) may describe physical conditions that may affect the results of a set of spectroscopic measurements.

[0066] Additionally or alternatively, the control device (210) may have the first spectrometer (220) perform the first part of the set of spectroscopic measurements on the first part of the unknown set, and the second spectrometer (220) perform the second part of the set of spectroscopic measurements on the second part of the unknown set. In this way, the control device (210) can reduce the amount of time required to perform a series of spectroscopic measurements in relation to having all spectroscopic measurements performed by a single spectrometer (220).

[0067] In this way, the control device (210) receives information identifying the results of a set of spectroscopic measurements performed on an unknown sample.

[0068] As further illustrated in FIG. 6, the process (600) may include a step (block 620) of determining whether an unknown sample corresponds to a quantification model. For example, the control device (210) may attempt to determine (e.g., using a processor (320), memory (330), storage component (340), etc.) whether the unknown sample is a substance configured to be quantified by the quantification model and / or whether the unknown sample contains within the substance a component configured to be quantified by the quantification model.

[0069] In some embodiments, the control unit (210) may use an SC-SVM classifier technique to determine whether an unknown spectrum corresponds to a quantification model. For example, the control unit (210) may determine a configurable parameter value (nu) to use the SC-SVM technique. The parameter value may correspond to the ratio of training set samples determined to be within the decision boundary to training set samples determined not to be within the decision boundary of the SC-SVM technique. In some embodiments, the control unit (210) may determine a decision boundary based on the parameter value. In some embodiments, the control unit (210) may set multiple possible decision boundaries using a cross-validation procedure and combine the results using multiple possible decision boundaries (e.g., through averaging) to determine whether an unknown spectrum corresponds to a quantification model.

[0070] In some embodiments, based on setting a judgment boundary to satisfy a parameter value (e.g., a parameter value of 0.5 that sets the judgment value such that half of the measurements in the training set are located within the judgment boundary and half of the measurements in the training set are located outside the judgment boundary), the control device (210) may determine a judgment value that corresponds to a distance metric from the measurements to the judgment boundary. For example, the control device (210) may determine a location on a set of axes of the spectrum of an unknown sample and determine the distance between this location and the nearest point of the judgment boundary. While some embodiments described herein are described with a graph or a set of axes, embodiments described herein may be determined without using a graph or a set of axes, for example, by using other representations of data related to the unknown spectrum.

[0071] In some embodiments, the control device (210) may determine a judgment value of an unknown spectrum. For example, the control device (210) may determine a distance from the unknown spectrum to a judgment boundary. In some embodiments, the control device (210) may determine a judgment value of another measurement located outside the judgment boundary. In this case, the control device (210) may determine a statistical metric to indicate the amount of standard deviation of the judgment value of the unknown spectrum with respect to the judgment value of another measurement outside the judgment boundary. For example, the control device (210) may determine a log-normal standard deviation based on a log-normal distribution and determine whether the standard deviation satisfies a threshold value (e.g., one standard deviation, two standard deviations, three standard deviations, etc.). In such cases, based on the fact that the measurement of the spectrum of the unknown sample is greater than a threshold amount of standard deviation from the judgment boundary (e.g., three standard deviations from the judgment boundary) for other measurements outside the judgment boundary, the control device (210) may determine that the unknown sample does not correspond to the quantification model (Block 620 - No). Alternatively, based on the fact that the measurement is less than a threshold amount of standard deviation from the judgment boundary, the control device (210) may determine that the unknown sample corresponds to the quantification model (Block 620 - Yes). Although this specification describes specific statistical techniques and / or specific threshold amounts of standard deviation, other statistical techniques and / or threshold amounts may be used.

[0072] In this way, the control device (210) enables the identification of an unknown spectrum that differs by a threshold amount from a quantification model without training the quantification model using a sample similar to the unknown sample (e.g., also differing by a threshold amount from a training set sample of the substance of interest). Additionally, the control device (210) reduces the cost, time, and utilization of computing resources (e.g., processing resources and memory resources) for acquiring, storing, and processing other samples to ensure accurate identification of a sample that differs by a threshold amount from the substance of interest and / or its concentration by reducing the amount of samples to be collected to generate the quantification model.

[0073] In this way, the control device (210) determines whether the unknown sample corresponds to the quantification model.

[0074] As further illustrated in FIG. 6, based on the determination that an unknown sample corresponds to a quantification model (Block 620 - Example), the process (600) may include the step of performing one or more spectroscopic decisions based on the results of a set of spectroscopic measurements (Block 630). For example, the control device (210) may perform one or more spectroscopic decisions based on the results of a set of spectroscopic measurements (e.g., using a processor (320), memory (330), storage component (340), etc.). In some embodiments, the control device (210) may assign the unknown sample to a specific class (e.g., representing a specific concentration of a set of concentrations of a component in the substance of interest).

[0075] In some embodiments, the control device (210) may assign a specific sample based on a reliability metric. For example, the control device (210) may determine, based on a quantification model, the probability that a specific spectrum is associated with each class of the quantification model (e.g., each candidate concentration). In this case, the control device (210) may assign an unknown sample to a class (e.g., a specific concentration) based on the fact that the specific probability of the class exceeds the other probability associated with the class. In this way, the control device (210) quantifies the sample by determining the concentration of a component in the substance of interest associated with the sample.

[0076] In some embodiments, for example, to perform local modeling at the original location when the quantification model exceeds a threshold of class, the control unit (210) may generate a local quantification model based on a first decision. The local quantification model may refer to a local quantification model generated using an SVM decision technique (e.g., kernel functions such as SVM-rbf, SVM-linear, etc.; probability value-based SVM, decision value-based SVM, etc.) based on a reliability metric associated with the first decision.

[0077] In some embodiments, the control device (210) may generate a local quantification model based on a first decision made using a global classification model. For example, if the control device (210) is used to determine the concentration of a component in an unknown sample, and multiple unknown samples are associated with different quantification models to determine the concentration of the component, the control device (210) may generate a subset of classes as local classes of the unknown sample using the first decision and generate a local quantification model associated with the local classes of the unknown sample. In this way, the control device (210) utilizes a hierarchical determination and quantification model to enhance spectroscopic classification. In this case, the control device (210) may determine whether the unknown sample corresponds to the local quantification model based on determining the distance metric of the unknown sample for a subset of different measurements of the local quantification model.

[0078] For example, when identifying raw materials to determine the concentration of a specific chemical substance in a plant material that is associated with a number of quantification models (e.g., whether the plant is grown indoors or outdoors, or whether it is grown in winter or summer), the control device (210) may perform a series of classification decisions to identify a specific quantification model. In this case, based on performing the series of decisions, the control device (210) may determine that the plant is grown indoors in winter and select a quantification model associated with the plant being grown indoors in winter to determine the concentration of a specific chemical substance. Based on the selection of the quantification model, the control device (210) may determine that an unknown sample corresponds to the quantification model and may quantify the unknown sample using the quantification model.

[0079] In some embodiments, the control device (210) may not be able to quantify an unknown sample using a quantification model. For example, based on one or more judgment values ​​or other reliability metrics that do not satisfy a threshold, the control device (210) may determine that the unknown sample cannot be accurately quantified using the quantification model (Block 630-A). Alternatively, the control device (210) may successfully quantify the unknown sample based on one or more judgment values ​​or other reliability metrics satisfying a threshold (Block 630-B).

[0080] In this way, the control device (210) performs one or more spectroscopic decisions based on the results of a set of spectroscopic measurements.

[0081] As further illustrated in FIG. 6, based on a determination that an unknown sample does not correspond to a quantification model (Block 620 - No) or based on a determination that one or more spectroscopic determinations failed (Block 630 - A), the process (600) may include the step of providing an output indicating that the unknown sample does not correspond to a quantification model (Block 640). For example, the control unit (210) may provide information indicating that the unknown sample does not correspond to a quantification model, for example, through a user interface (e.g., using a processor (320), memory (330), storage component (340), communication interface (370), etc.). In some embodiments, the control unit (210) may provide information associated with identifying the unknown sample. For example, based on an attempt to quantify the amount of a specific chemical substance in a specific plant, and based on a determination that an unknown sample is not a specific plant (but instead is a different plant, for example, based on human error), the control device (210) may provide information to identify a different plant. In some embodiments, the control device (210) may obtain another quantification model and use the other quantification model to identify an unknown sample based on a determination that the unknown sample does not correspond to the quantification model.

[0082] In this way, the control device (210) reduces the possibility of providing inaccurate information based on the false positive identification of an unknown sample and enables the technician to correct the error by providing information that helps determine, for example, that the unknown sample is a different plant than a specific plant.

[0083] In this way, the control device (210) provides an output indicating that the unknown sample does not correspond to the quantification model.

[0084] As further illustrated in FIG. 6, when performing one or more spectroscopic determinations, based on the success of classification (block 630-B), the process (600) may include a step (block 650) of providing information identifying a classification regarding an unknown sample. For example, the control device (210) may provide information identifying a quantification regarding an unknown sample (e.g., using a processor (320), memory (330), storage component (340), communication interface (370), etc.). In some embodiments, the control device (210) may provide information identifying a specific class of the unknown sample. For example, the control device (210) may identify the unknown sample by providing information indicating that a specific spectrum associated with the unknown sample has been determined to be associated with a specific class corresponding to a specific concentration of a component in the substance of interest.

[0085] In some embodiments, the control device (210) may provide information indicating a confidence metric associated with assigning an unknown sample to a specific class. For example, the control device (210) may provide information identifying the probability that an unknown sample is associated with a specific class, etc. In this way, the control device (210) provides information indicating the probability that a specific spectrum is accurately assigned to a specific class.

[0086] In some embodiments, the control device (210) may provide quantification based on having performed a series of classifications. For example, based on identifying a local quantification model associated with the class of an unknown sample, the control device (210) may provide information identifying the concentration of a substance species in the unknown sample. In some embodiments, the control device (210) may update the quantification model based on having performed a series of quantifications. For example, the control device (210) may generate a new quantification model containing the unknown sample as a sample in the training set based on determining the quantification of the unknown sample to a specific concentration of a component in the substance of interest.

[0087] In this way, the control device (210) provides information for identifying an unknown sample.

[0088] FIG. 6 illustrates an exemplary block of process (600), but in some embodiments, process (600) may include additional blocks, a smaller number of blocks, different blocks, or blocks arranged differently compared to that illustrated in FIG. 6. Additionally or alternatively, two or more of the blocks of process (600) may be performed in parallel.

[0089] FIGS. 7A and 7B are drawings illustrating an exemplary embodiment (700) regarding a prediction success rate associated with the exemplary process (600) illustrated in FIG. 6. FIGS. 7A and 7B illustrate exemplary results of identifying raw materials using a hierarchical support vector machine (hier-SVM-linear) based technique.

[0090] As illustrated in FIG. 7a and reference number (705), the control unit (210) may cause the spectrometer (220) to perform a series of spectroscopic measurements. For example, the control unit (210) may provide a command for the spectrometer (220) to acquire the spectrum of an unknown sample to determine the concentration of a component in the unknown sample. As illustrated in reference number (710) and reference number (715), the spectrometer (220) may receive an unknown sample and perform a series of spectroscopic measurements on the unknown sample. As illustrated in reference number (720), the control unit (210) may receive the spectrum of the unknown sample-based spectrometer (220) that has performed a series of spectroscopic measurements on the unknown sample.

[0091] As illustrated in FIG. 7b, the control device (210) can perform spectroscopic quantification using a quantification model (725). The quantification model (725) includes a single class (730) determined based on a parameter value (nu) such that the judgment boundary of a single class (730) results in a threshold ratio of samples in the training set that are within the judgment boundary for samples in the training set that are not within the judgment boundary. In this case, the quantification model (725) may be associated with multiple sub-classes corresponding to multiple different concentrations of a component in samples in the training set. As illustrated by reference number (735) and reference number (740), the spectrum of an unknown sample is determined not to correspond to a quantification model based on a standard deviation value (e.g., σ = 3.2) for the distance of the sample to the judgment boundary satisfying a threshold value (e.g., 3). As illustrated by reference number (745), the control device (210) provides an output to the client device (750) indicating that the unknown sample does not correspond to the quantification model, rather than providing a false positive identification of the unknown sample as a specific concentration of a component in the substance of interest.

[0092] As previously mentioned, FIGS. 7a and 7b are provided merely as examples. Other examples are possible and may differ from those described in relation to FIGS. 7a and 7b.

[0093] In this way, the control device (210) reduces the possibility of providing inaccurate spectroscopic results by avoiding the possibility of the quantification model being trained to identify an unknown sample as a specific concentration of a component in the substance of interest.

[0094] The foregoing provides examples and descriptions, but is not intended to limit implementation to the exact form disclosed or to present all embodiments. Changes and variations may be possible in light of the foregoing or may be obtained by carrying out the implementation.

[0095] Some embodiments are described herein with respect to thresholds. As used herein, satisfying a threshold may mean that a value is greater than the threshold, greater than the threshold, higher than the threshold, greater than the threshold, less than the threshold, smaller than the threshold, lower than the threshold, less than or equal to the threshold, or equal to the threshold.

[0096] It is evident that the systems and / or methods described herein may be implemented in various forms of hardware, firmware, or combinations of hardware and software. The actual specialized control hardware or software code used to implement such systems and / or methods is not intended to limit the implementation of the invention. Accordingly, the operation and behavior of the systems and / or methods have been described herein without reference to specific software code, and it is understood that the software and hardware may be designed to implement the systems and / or methods based on the description in this specification.

[0097] Specific combinations of features are described in the claims and / or disclosed in the specification, but these combinations are not intended to limit the disclosure of possible implementations. In fact, many of these features may be combined in ways not specifically described in the claims and / or disclosed in the specification. Each dependent claim listed below may directly depend on only one claim, but the content of possible implementations includes each dependent claim citing all other claims within the claims.

[0098] Any element, action, or command used herein shall not be interpreted as being decisive or essential to the invention unless explicitly described as such. Additionally, the singular used herein is intended to include one or more items and may be used interchangeably with "one or more." Furthermore, the terms "series" and "set" used herein are intended to include one or more items (e.g., related items, unrelated items, and combinations of related and unrelated items, etc.) and may be used interchangeably with "one or more." If only a single item is intended, the term "one" or similar language is used. Additionally, terms such as "have," "having," and "having" used herein are intended to be open-ended terms. Furthermore, the phrase "based on" is intended to mean "at least partially based on" unless explicitly stated otherwise.

Claims

Claim 1 A method comprising: a step of providing a command to a spectrometer for performing one or more spectroscopic measurements on a set of samples including samples of a plurality of concentrations; a step of receiving a set of spectra for the set of samples from the spectrometer based on the command provided by the device for performing one or more spectroscopic measurements; a step of generating a global classification model and a plurality of local quantification models for identifying a type of substance based on the device receiving the set of spectra for the set of samples, wherein the plurality of local quantification models include a first local quantification model for determining the concentration of a component in a first type of substance and a second local quantification model for determining the concentration of the component in a second type of substance; and a step of providing a plurality of quantification models including the global classification model and the plurality of local quantification models. Claim 2 A method according to claim 1, wherein the step of providing the plurality of quantification models comprises the step of providing the plurality of quantification models to a first different device associated with a first different spectrometer and a second different device associated with a second different spectrometer. Claim 3 A method according to claim 1, further comprising the step of training and validating a plurality of local quantification models associated with the global classification model. Claim 4 A method according to claim 1, wherein the plurality of quantification models comprises: a first quantification model corresponding to a first class of the global classification model; and a second quantification model corresponding to a second class of the global classification model. Claim 5 A method according to claim 1, wherein the plurality of quantification models comprises: a first quantification model corresponding to a first class of a local classification model generated based on the global classification model; and a second quantification model corresponding to a second class of the local classification model. Claim 6 A method according to claim 1, wherein the command for performing one or more spectroscopic measurements comprises a command for performing one or more spectroscopic measurements for a training set and a validation set. Claim 7 A method according to claim 1, wherein the plurality of quantification models comprises: a first quantification model for determining the concentration of a specific protein in a first type of meat identified using in-situ local modeling; and a second quantification model for determining the concentration of the specific protein in a second type of meat identified using in-situ local modeling. Claim 8 A device comprising: one or more memories; and one or more processors coupled to the one or more memories, wherein the one or more processors receive a set of spectra for a set of samples associated with one or more spectroscopic measurements for a set of samples including samples of a plurality of concentrations from a spectrometer; and, based on receiving the set of spectra for a set of samples, generate a global classification model for identifying a type of substance and a plurality of local quantification models, wherein the plurality of local quantification models include a first local quantification model for determining the concentration of a component in a first type of substance and a second local quantification model for determining the concentration of the component in a second type of substance; and configured to provide a plurality of quantification models including the global classification model and the plurality of local quantification models. Claim 9 In claim 8, the one or more processors providing the plurality of quantification models are configured to provide the plurality of quantification models to a first different device associated with a first different spectrometer and a second different device associated with a second different spectrometer. Claim 10 In claim 8, the device wherein the one or more processors are configured to train a plurality of local quantification models associated with the global classification model. Claim 11 In claim 8, the apparatus comprises a plurality of quantification models, wherein the plurality of quantification models includes a first quantification model corresponding to a first class of the global classification model; and a second quantification model corresponding to a second class of the global classification model. Claim 12 In claim 8, the apparatus comprises a plurality of quantification models, a first quantification model corresponding to a first class of a local classification model generated based on the global classification model; and a second quantification model corresponding to a second class of the local classification model. Claim 13 In claim 8, the apparatus comprises a plurality of quantification models, a first quantification model for determining the concentration of a specific protein in a first type of meat identified using local modeling at the original location; and a second quantification model for determining the concentration of the specific protein in a second type of meat identified using local modeling at the original location. Claim 14 A non-transient computer-readable medium for storing a set of instructions, wherein the set of instructions comprises, when executed by one or more processors of the device, an operation of providing a command for the device to perform one or more spectroscopic measurements on a set of samples including samples of a plurality of concentrations using a spectrometer; an operation of receiving a set of spectra for the set of samples from the spectrometer based on providing the command for performing one or more spectroscopic measurements; an operation of generating a global classification model and a plurality of local quantification models for identifying a type of substance based on receiving the set of spectra for the set of samples, wherein the plurality of local quantification models include a first local quantification model for determining the concentration of a component in a first type of substance and a second local quantification model for determining the concentration of the component in a second type of substance; and one or more instructions for causing the device to perform an operation of providing a plurality of quantification models including the global classification model and the plurality of local quantification models. Claim 15 In paragraph 14, the one or more instructions that cause the device to provide the plurality of quantification models are a non-transient computer-readable medium that causes the device to provide the plurality of quantification models to a first different device associated with a first different spectrometer and a second different device associated with a second different spectrometer. Claim 16 In paragraph 14, the above one or more commands are a non-transient computer-readable medium that enables the device to verify a plurality of local quantification models associated with the global classification model. Claim 17 A non-transient computer-readable medium, wherein, in paragraph 14, the plurality of quantification models comprises: a first quantification model corresponding to a first class of the global classification model; and a second quantification model corresponding to a second class of the global classification model. Claim 18 A non-transient computer-readable medium according to claim 14, wherein the plurality of quantification models comprises: a first quantification model corresponding to a first class of a local classification model generated based on the global classification model; and a second quantification model corresponding to a second class of the local classification model. Claim 19 In paragraph 14, a non-transient computer-readable medium comprising a command for performing one or more spectroscopic measurements, wherein the command for performing one or more spectroscopic measurements for a training set and a validation set. Claim 20 In claim 14, the plurality of quantification models comprises: a first quantification model for determining the concentration of a specific protein in a first type of meat identified using in-situ local modeling; and a second quantification model for determining the concentration of the specific protein in a second type of meat identified using in-situ local modeling, in a non-transient computer-readable medium. Claim 21 The method of claim 1, wherein the device receives information identifying the result of a spectroscopic measurement performed on an unknown sample, determines a determination boundary for each of the plurality of quantification models, determines a distance metric for the determination boundary of the spectroscopic measurement value performed on the unknown sample with respect to each quantification model, and determines whether the spectroscopic measurement value performed on the unknown sample corresponds to each of the plurality of quantification models based on the distance metric.

Images