Systems and methods for spectrum-based disease detection

Abstract

A system includes a light source that emits broadband light through a test cartridge containing a fluid. After the broadband light passes through the test cartridge, the spectrometer disperses the broadband light onto a plurality of pixel detectors of the detector array. A digitizer reads the array of raw spectral values from the array of detectors. The signal processing circuitry deconvolves the original array of spectral values based on the spectral response of each of the plurality of pixel detectors to generate a deconvolved array of spectral values; feeding the array of deconvolution spectral values into a trained machine learning model to obtain an indication of the presence of one or more components in the fluid; and outputting an indication. The fluid may be human saliva or another biological fluid obtained from a patient. The indication may be used to diagnose a patient with a disease or condition.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. Provisional Patent Application No. 63 / 325,076, filed on March 29, 2022, the entire contents of which are incorporated herein by reference. BRIEF DESCRIPTION OF THE DRAWINGS

[0003] Figure 1 is a block diagram of a system for spectroscopy-based disease detection according to an embodiment.

[0004] Figure 2 is the implementation scheme Figure 1 A system is similar to the block diagram of the system except that it uses two spectrometers instead of one.

[0005] Figure 3 A method for processing a first raw spectral value array and a second raw spectral value array in an embodiment is shown.

[0006] Figure 4 A method for processing a first raw spectral value array and a second raw spectral value array in an embodiment is shown.

[0007] Figure 5 shows how an embodiment can be implemented using fiber optic components Figure 2 system.

[0008] Figure 6 An embodiment of the present invention can be implemented using a fiber optic component. Figure 2 Another way of system.

[0009] Figure 7 is the implementation scheme Figure 2 A system is similar to the block diagram of the system except that it uses two separate beams instead of one.

[0010] Figure 8 Shows Figure 1 How can each pixel detector of the discrete sampling process of the continuous spectral density p(λ) effectively combine the spectral density p(λ) with the sensitivity function f unique to each pixel detector? i (λ) or kernel for convolution.

[0011] Fig. 9 shows a wavelength λ j How light contributes to several spectral values.

[0012] Fig.10 It is the implementation plan Figure 1 A functional diagram of a computing system with one example of signal processing circuitry. DETAILED DESCRIPTION

[0013] A biomarker is a measurable indicator of a disease or physical condition in an organism. A physical condition can be a normal biological process, a pathogenic process, or a response to a therapeutic intervention (e.g., a pharmacological response to a prescribed drug). For clinical purposes, biomarkers can be used to guide or narrow down a patient's treatment options. More specifically, biomarkers can be used to predict (i.e., predict a patient's clinical outcome), diagnose (i.e., help diagnose a patient), or prognose (i.e., determine the overall outcome).

[0014] The spread of SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2, also known as Covid-19 disease) has renewed interest in improving tests that can detect this and other diseases. Currently, the most accurate diagnosis of the SARS-Cov-2 virus uses polymerase chain reaction (PCR), such as quantitative reverse transcription PCR (RT-qPCR), which amplifies DNA and RNA sequences to make them more easily detectable. Nasal swab testing using PCR-based testing is accurate, but there are several limitations, including the way samples are handled (e.g., improper swabbing and storage), the requirement that sampling be performed during the acute phase, and long testing times. For example, PCR collection can take 2 to 4 hours, and the overall handling and processing can take more than 12 hours. PCR machines are also bulky, expensive, and require skilled personnel to operate correctly.

[0015] Antigen tests are also now commonly used to detect SARS-Cov-2. Antigen tests identify the presence of the virus in nasal and throat secretions by looking for proteins produced by the virus (as opposed to diagnostic tests that look directly for genetic material). Advantageously, antigen tests take only 15 minutes, are inexpensive, and can be performed at home without the need for a medical professional or expensive equipment. However, antigen tests do not have the accuracy of PCR-based tests and are known for high false-negative rates, especially in patients with low viral loads. Antigen tests can also give false results due to improper handling (e.g., insufficient swabbing). They also require reagents, which can be difficult to produce and obtain during an outbreak.

[0016] Recently, light-based diagnostic technologies are being explored to combine the sensitivity and specificity of PCR-based tests with the low cost, high speed, and scalability of antigen tests. Some of these light-based tests require no reagents, thereby eliminating important issues with PCR and antigen-based tests. These light-based tests perform spectroscopic analysis (e.g., attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy) on samples obtained from nasal swabs or mouthwash to identify spectral features known to be associated with the presence of COVID-19.

[0017] The present disclosure includes embodiments for spectroscopically detecting the presence of one or more components in a fluid sample. Specifically, the fluid sample is first measured to obtain an absorption spectrum. The absorption spectrum is then input into a machine learning model that is trained to identify one or more features in the absorption spectrum. Each feature is considered to be unique to the corresponding component, and therefore the presence of a feature in the absorption spectrum indicates the presence of the corresponding component in the fluid sample. The machine learning model outputs an indication (e.g., a probability) of such presence.

[0018] Some of the present embodiments can be used with biological fluids (e.g., blood, saliva, mucus, etc.) to quickly diagnose the presence of diseases in fluids. For example, when used with a small amount of saliva samples, some of these embodiments can detect SARS-CoV-2 viral infection within thirty seconds. Here, "a small amount" means that a typical human patient can provide a sample non-invasively with very little workload (e.g., spitting into a cup) and minimal help (if any). More generally, these embodiments can be applicable to detecting any type of chemical or cellular biomarkers obtained from any type of organism. These biomarkers indicate the disease state or physiological state of an organism. Examples of such states include, but are not limited to, viral infections, bacterial infections, toxin-based diseases, cancer, heart disease, or the lack of these diseases.

[0019] Medical workers (e.g., doctors, nurses, etc.) can use the instructions output to diagnose human patients with diseases or illnesses associated with biomarkers. However, medical workers can also diagnose human patients lacking the disease or illness based on the instructions. Therefore, some of the present embodiments can not only be used to identify whether the patient is ill, but also can be used to determine whether the patient is disease-free. In either case, medical workers can then provide appropriate therapeutic intervention as required. For example, if the medical worker diagnoses that the patient suffers from a disease, he or she can prescribe, provide or perform therapeutic intervention to treat the disease. Medical workers can use additional information and possible additionally measured biomarkers to determine therapeutic intervention (e.g., age, sex, blood pressure, body temperature, etc.). The example of therapeutic intervention includes but is not limited to surgical operation, non-surgical medical operation and one or more drug prescriptions.

[0020] Some of the present embodiments can operate without the use of chemical reagents, which helps to speed up and simplify sample preparation. Simplifying sample preparation improves accuracy by reducing errors that may occur during processing (e.g., collecting samples and injecting them into the system). Eliminating chemical reagents also reduces the threat of supply chain disruptions that may occur during epidemic response. In addition, by updating the machine learning model, these embodiments can be quickly and easily adapted (e.g., via software updates) to detect different diseases or physiological states (e.g., mutations of the virus).

[0021] The combination of speed and lack of reagent makes at least some of the present embodiments very suitable for real-time removal applications. For example, some of the present embodiments can be used for screening airline passengers before boarding or entering an airport terminal. In related applications, some of the present embodiments can be used for testing people who wish to enter the interior of a building (e.g., a triage center, a medical institution, a laboratory, an arena, a stadium, etc.). However, these embodiments can also be used for testing other people in other cases, such as students who are sick during the day, patients who have been admitted to a hospital or clinic, and individuals who visit outpatient medical institutions (e.g., a doctor's office).

[0022] Some of the present embodiments include or use one or more spectrometers, each spectrometer disperses broadband light onto a plurality of pixel detectors forming a detector array. A wavelength-dependent sensitivity function is associated with each pixel detector, and the wavelength-dependent sensitivity function peaks at a center wavelength. Thus, although each pixel detector detects a small portion of the wavelength range of broadband light (e.g., 0.1 nm to 1 nm), for simplicity, its electrical output is associated with its center wavelength. Thus, a plurality of center wavelengths are associated with the detector array, and the plurality of center wavelengths form a one-to-one correspondence with the plurality of pixel detectors.

[0023] The machine learning model accepts multiple spectral values ​​as input. Each of these spectral values ​​is associated with a corresponding reference wavelength. However, the machine learning model will likely have been trained using a reference wavelength that is different from the central wavelength of a particular spectrometer. These differences arise from manufacturing tolerances during the construction of the spectrometer. Therefore, simply feeding the output of the pixel detector directly into the machine learning model may reduce accuracy.

[0024] Some of the present embodiments use a technique referred to herein as wavelength shift deconvolution to solve this problem. Similar to the deconvolution techniques of the prior art, wavelength shift deconvolution sharpens the measured spectrum by eliminating the blurring effect of the sensitivity function. However, unlike the deconvolution techniques of the prior art, wavelength shift deconvolution corrects the input spectral values ​​for the difference between the center wavelength and the reference wavelength. Therefore, wavelength shift deconvolution improves the accuracy indicated by the machine learning model output by ensuring that the spectral values ​​fed into the machine learning model correctly indicate the measured values ​​at the reference wavelength. In addition, wavelength shift deconvolution allows the same machine learning model to be used across all units, regardless of changes in the center wavelength. Wavelength shift deconvolution can also be used to change the number of spectral values ​​to match the number of inputs to the machine learning model. Therefore, for wavelength shift deconvolution, the number of inputs does not need to match the number of pixel detectors.

[0025] Figure 11 is a block diagram of a system 100 for spectroscopy-based disease detection. The system 100 includes a light source 102 that emits broadband light 103 through a fluid 108 confined within a test cartridge 106. The portion of the broadband light 103 that exits the test cartridge 106 is referred to as transmitted light 105. The system 100 also includes a cartridge holder 104 that is shaped to receive the test cartridge 106 and position the test cartridge 106 such that the broadband light 103 enters the fluid 108 and the transmitted light 105 exits the fluid 108.

[0026] To enhance the transmission of the broadband light 103 through the test box 106, the box holder 104 may include a pair of relatively optically transparent windows through which the broadband light 103 enters and exits the test box 106. Alternatively, the box holder 104 may form an aperture (e.g., a hole) for the broadband light 103 to pass through. Similarly, the test box 106 may be made of a material that is at least partially transparent to the broadband light 103. For example, the test box 106 may be a glass cuvette, in which case the box holder 104 may be a cuvette holder. However, the test box 106 may also be made of another transparent material (e.g., plastic, crystal) without departing from the scope thereof. As described below with respect to Figure 7 As discussed, the test cartridge 106 may be made of multiple materials rather than a single uniform material.

[0027] The system 100 also includes a spectrometer 110 that measures the absorption spectrum of the fluid 108. The fluid 108 includes one or more components, each of which has an absorption spectrum that is considered unique and can therefore be used to identify the presence of the component in the fluid 108. The characteristics of the absorption spectrum of a component that make it unique are collectively referred to as "signatures". Therefore, some of the present embodiments identify the presence of a component by identifying its signature in the measured absorption spectrum. The properties that contribute to the signature can include the presence of one or more absorption dips, the central wavelength of these absorption dips, the width of such absorption dips (e.g., full width at half maximum), and the slope of the spectrum within certain spectral bands (e.g., rising, falling, or flat).

[0028] Examples of components in fluid 108 include microorganisms, such as archaea, bacteria, prokaryotes, and eukaryotes. When system 100 measures these organisms, they do not need to be alive. Other components include viruses, prions, microbial fragments (e.g., organelles), macromolecules (e.g., proteins, DNA, nucleic acids, carbohydrates, etc.), and organic compounds (e.g., small molecules and polymers). Components may be produced by one or more microorganisms (e.g., viral proteins) present or previously present in fluid 108. Alternatively, the host may produce components in response to the presence of microorganisms (e.g., antibodies) in the host. Fluid 108 may alternatively or additionally contain other components without departing from the scope thereof.

[0029] The fluid 108 may also be produced by dissolving a solid in a liquid solute or by mixing a solid with a base liquid. In some embodiments, the fluid 108 is replaced with a solid (e.g., a tissue sample) that is at least partially transparent to the broadband light 103. Thus, the present embodiments are not limited to fluids, i.e., some of the present embodiments may use gases instead.

[0030] Fluid 108 can be a sample obtained from a human patient. Examples of such samples include saliva, blood, mucus, and urine. Alternatively, the sample can be taken from a non-human animal, such as a domestic pet, a farm animal, or a zoo animal. However, fluid 108 can be another biological fluid without departing from the scope of this invention. In addition, fluid 108 does not necessarily need to be taken from humans or animals. For example, fluid 108 can be a sample taken from a body of water (e.g., a lake, a reservoir, an ocean, etc.), a well, a water treatment facility, a sewage treatment facility, or a sewer system.

[0031] In order to identify the features of interest in the absorption spectrum, the broadband light 103 may cover a sufficiently wide wavelength range to ensure that all the characteristics required to determine the features are measured. Similarly, the spectrometer 110 may be designed to operate within this wavelength range. In some embodiments, the wavelength range of the broadband light 103 spans at least a portion of the visible region of the electromagnetic spectrum (e.g., approximately 350nm to 750nm in air or vacuum). In other embodiments, the wavelength range spans at least a portion of the infrared region of the electromagnetic spectrum. For example, the wavelength range may span the near infrared region of 750nm to 1400nm. Alternatively, the wavelength range may span 900nm to 1700nm. In some embodiments, the wavelength spans a portion of both the visible and infrared regions (e.g., 600nm to 1000nm). The wavelength range may at least partially span one or more other regions of the electromagnetic spectrum (e.g., ultraviolet, mid-infrared, far-infrared, etc.) without departing from its scope. Due to technical reasons described in detail below, spanning more than twice the wavelength range may be best detected using two independent spectrometers operating in parallel. However, the wavelength range of the broadband light 103 is not so limited. Thus, in some embodiments, the wavelength range spans more than twice (eg, 350 nm to 1700 nm).

[0032] The light source 102 uses one or more light emitting devices to create the broadband light 103 as a single beam. The broadband light 103 may be incoherent, in which case the light source 102 may include a lamp, a discharge tube, or an LED source. The incoherent broadband light 103 may be white light (or light of similar color) containing all wavelengths in the wavelength range. Alternatively, the broadband light 103 may be coherent, in which case the light source 102 may include a laser. The coherent broadband light 103 may be a quasi-white light (or similar quasi-colored light) having discrete frequency components with equal frequency intervals. This quasi-white light may be generated by, for example, an ultrafast pulsed laser (e.g., a mode-locked femtosecond titanium sapphire laser or a fiber laser). The wavelength range of the quasi-white light may be further extended using nonlinear optical techniques, such as four-wave mixing in nonlinear crystals or optical media (e.g., photonic crystal fibers). As described in more detail below, the light source 102 may combine the outputs of two or more light emitting devices. The two or more light emitting devices may be an incoherent light source, a coherent light source, or a combination of a coherent light source and an incoherent light source.

[0033] The broadband light 103 exits the fluid 108 and the test box 106 as transmitted light 105, which is subsequently processed by a spectrometer 110. The spectrometer 110 includes a dispersive element 112 that disperses the transmitted light 105 onto a plurality of (N) pixel detectors 116(1)…116(N) arranged as a detector array 114, where N is an integer greater than 1. Each pixel detector 116(i) of the detector array 114 outputs a corresponding photocurrent 118(i) that is approximately linearly proportional to the irradiance of the portion of the transmitted light 105 incident on the pixel detector 116(i). Because each pixel detector 116(i) measures optical power, the spectrometer 110 may also be referred to as a spectroradiometer. Although the dispersive element 112 is Figure 1 1 is shown as a transmissive diffraction grating, but the dispersive element 112 may alternatively be a reflective diffraction grating, a prism, or another type of optical element that spatially disperses polychromatic light. The spectrometer 110 may include Figure 1 One or more other components not shown in the figure, such as slits, lenses, mirrors (e.g., for beam steering), filters, etc. In other embodiments, spectrometer 110 includes a scanning monochromator. In these embodiments, only one photodetector is required.

[0034] The detector array 114 may be a one-dimensional array, such as Figure 1 Alternatively, the detector array 114 may be a two-dimensional array (e.g., a CMOS or CCD image sensor). In this case, the output of each row (or column) of pixels may be integrated to obtain a single value. Each pixel detector 116 (i) is based on a spectral sensitivity function f i(λ) detects the spectral sub-range of the transmitted light 105. The sensitivity function f i (λ) is also called kernel function, blur function or bandwidth function. Usually, each sensitivity function f i (λ) are all at the central wavelength The central wavelength is the wavelength at which the photocurrent 118(i) output by the pixel detector 116(i) is at a maximum. Therefore, the pixel detectors 116(1) ... 116(N) are connected to the multiple central wavelengths. The following is a one-to-one correspondence between the spectral sensitivity function f i Additional details of (λ).

[0035] Each pixel detector 116(i) Figure 1 114 are shown as photodiodes. Each photodiode can be made of a material that makes the photodiode highly responsive in the spectral subrange that it detects. Examples of such materials include silicon (Si), gallium arsenide (GaAs), indium gallium arsenide (InGaAs), and germanium (Ge). When detecting low light levels, each photodiode can be an avalanche photodiode. In this case, the detector array 114 can be a multi-pixel photon counter or a silicon multiplier. Each pixel detector 116 (i) can be another type of light detector, photodetector, or light sensor without departing from the scope thereof.

[0036] The spectrometer 110 further comprises a digitizer 120 which reads the raw spectral values ​​s1 . . . s from the detector array 114. N Specifically, the digitizer 120 samples each photocurrent 118 (i) and outputs the corresponding raw spectrum value s i , that is, the digital representation of the sample. Figure 1In an example of , the digitizer 120 includes a multiplexer (MUX) 122 that connects one of the photocurrents 118 to the input of an analog-to-digital converter (ADC) 124. In this case, the digitizer 120 digitizes the photocurrents 118 sequentially (i.e., one at a time). Alternatively, the digitizer 120 may include N ADCs, in which case each photocurrent 118(i) is digitized by its own dedicated ADC. While the dedicated ADCs allow all photocurrents 118(1)...118(N) to be processed simultaneously, there may be thousands of pixel detectors 116 (i.e., N>1000) or more, and having so many dedicated ADCs may be very expensive and complex to build. Alternatively, the digitizer 120 may include multiple ADCs (e.g., four ADCs integrated into a single quad ADC chip) and multiple multiplexers 122 that are wired so that each of the pixel detectors 116 is sampled by one of the multiple ADCs. The digitizer 120 may be alternatively configured without departing from the scope thereof. Figure 1 The digitizer 120 is shown as part of the spectrometer 110 , but the digitizer 120 may alternatively be separate from the spectrometer 110 .

[0037] exist Figure 1 In the array 126, the raw spectral values ​​s1...s N The order of matches the order of pixel detectors 116(1)...116(N) and thus the center wavelength Maintaining this order can simplify subsequent signal processing. However, for the original spectral values ​​s1...s N There is no requirement for the order, as long as the order is known, the subsequent signal processing can be implemented correctly.

[0038] The spectrometer 110 can also be operated as a spectrophotometer. Specifically, the system 100 can be operated twice: once with the fluid 108 in the test cartridge 106 and once without it. The measurements obtained without the fluid 108 are used as a baseline, which can be subtracted from the measurements obtained with the fluid 108 to offset system effects that are not caused by the fluid 108 itself. Such system effects include non-uniform absorption spectra of the test cartridge 106, the cartridge holder 104, and other optical devices between the light source 102 and the spectrometer 110. Another system effect is the non-uniformity of the broadband light 103 (i.e., the broadband light 103 may not be completely white).

[0039] System 100 also includes signal processing circuitry 130 that processes the spectral sensitivity function f of each of pixel detectors 116 based on the spectral sensitivity function f i (λ) for the original spectral values ​​s1...s NThe array 126 is deconvolved to generate a deconvolved spectral value As described in more detail below, the spectral sensitivity function f1(λ)...f N (λ) may be combined into a kernel matrix, which is multiplied by array 126 to obtain array 132.

[0040] The signal processing circuit 130 also deconvolutes the spectral value The array 132 is fed into a machine learning model 140, which transforms the array 132 into an indication 142 of the presence of one or more components in the fluid 108. The indication 142 may include, for example, a probability of the presence (or absence) of a particular component in the fluid 108. The machine learning model 140 may determine the probability by comparing the array 132 with known features of the particular component. Thus, the probability quantifies the likelihood that the feature appears in the array 132. In some embodiments, the indication 142 includes a plurality of probabilities, each probability indicating the presence (or absence) of a corresponding component in a plurality of components in the fluid 108.

[0041] The machine learning model 140 can be any type of machine learning model known in the art. Examples include, but are not limited to, artificial neural networks, support vector machines, decision trees, random forests, regression models, and combinations thereof. The machine learning model 140 is trained to recognize one or more components using supervised data in which known features of the one or more components are labeled. In this sense, the feature is "known" to the machine learning model 140.

[0042] In some embodiments, the machine learning model 140 is stored in and executed by the signal processing circuit 130 or another circuit of the system 100. In other embodiments, the machine learning model 140 is provided by a third party and is therefore not part of the system 100. For example, the machine learning model 140 may be implemented on a remote computer server operated by a third party (e.g., platform as a service). In these embodiments, the signal processing circuit 130 (or another component of the system 100) transmits the array 132 (e.g., via the Internet) to the remote computer server for processing using the machine learning model 140. The system 100 then receives an indication 142 from the remote computer server. Using a remote server may be advantageous when the computing resources required to execute the machine learning model 140 exceed the resources that the signal processing circuit 130 can provide.

[0043] Figure 2 is with Figure 11 is a block diagram of a system 200 that is similar to system 100, except that two spectrometers are used instead of one. Specifically, system 200 includes a beam splitter 230 that divides transmitted light 105 into a first portion 205(1) and a second portion 205(2). System 200 includes a first spectrometer 210(1) that processes first portion 205(1) and a second spectrometer 210(2) that processes second portion 205(2). Each of spectrometers 210(1) and 210(2) is similar to Figure 1 Spectrometer 110 operates in accordance with and may therefore be configured in any manner described above for spectrometer 110. Spectrometers 210(1) and 210(2) may process respective portions 205(1) and 205(2) simultaneously.

[0044] The first spectrometer 210(1) includes a first dispersive element 212(1) that disperses the first portion 205(1) into a plurality of (N) pixel detectors 216(1) ... 216(N) that are arranged as a first detector array 214(1). Similarly, the second spectrometer 210(2) includes a second dispersive element 212(2) that disperses the second portion 205(2) into a plurality of (M) pixel detectors 218(1) ... 218(M) that are arranged as a second detector array 214(2), where M is an integer greater than 1. Furthermore, M may be equal to, less than, or greater than N. The dispersive elements 212(1) and 212(2) are similar to Figure 1 The dispersive element 112, the detector arrays 214(1) and 214(2) are similar to Figure 1 The detector array 114 and the pixel detectors 216 and 218 are similar to Figure 1 Pixel detector 116.

[0045] The analog output of the first detector array 214(1) is digitized to create raw spectral values ​​s1...s N Similarly, the analog output of the second detector array 214 (2) is digitized to create raw spectral values ​​t1 ... t M In order to make Figure 2 For clarity, the digitizer is not shown. Arrays 226(1) and 226(2) are then fed to Figure 1 The signal processing circuit 130 is used to Figure 1 The system 200 is based on both arrays 226 (1) and 226 (2) rather than just on Figure 1 An array 126 of generates indications 142 .

[0046] and Figure 1 Compared to system 100, system 200 advantageously improves accuracy when the wavelength range of broadband light 103 is particularly large. For example, it is known in the art that when the wavelength range of incident light is large, consecutive diffraction orders of the grating overlap. In this case, different wavelengths will be incident on the same pixel detector 116. System 200 solves this problem by assigning different wavelength ranges to spectrometers 210(1) and 210(2). Specifically, the first spectrometer 210(1) operates within a first spectrometer wavelength range, while the second spectrometer 210(2) operates within a second spectrometer wavelength range that is different from the first spectrometer wavelength range. For example, the first spectrometer 210(1) may operate using infrared light, while the second spectrometer 210(2) may operate using visible light. The wavelength range of broadband light 103 may span both the first spectrometer wavelength range and the second spectrometer wavelength range. The first spectrometer wavelength range and the second spectrometer wavelength range may overlap at least partially or not at all.

[0047] Figure 3 shows the process for processing the raw spectral values ​​s1...s N The first array 226 (1) and the original spectral values ​​t1 ... t M Method 300 generates a deconvolution array 340 which is fed to Figure 1 The method 300 may be performed by: Figure 1 The signal processing system 130 is used to perform.

[0048] In block 302 of method 300, arrays 226(1) and 226(2) are combined into a composite array 330 having N+M raw spectral values, denoted as c1...c N+M .like Figure 3 As shown, the composite array 330 can be generated by appending the second array 226(2) to the first array 226(1). Alternatively, the first array 226(1) can be appended to the second array 226(2). However, it is not required to maintain the original spectral values ​​s1...s in the composite array 330. N The order or raw spectral values ​​t1...t M Thus, composite array 330 may be generated by interleaving arrays 226(1) and 226(2) or by another merging technique. Composite array 330 may be generated by arranging the pixels of detector 216 in the order of their central wavelength. and the central wavelength of the pixel detector 218 However, composite array 330 may be sorted in another manner, or not sorted, without departing from the scope thereof.

[0049] In block 304, the composite array 330 is deconvolved to generate deconvolved spectral values The deconvolution array 340 of FIG. 340 can be used. Any deconvolution method known in the art can be used, such as frequency domain deconvolution (also known as Fourier deconvolution or Fourier self-deconvolution) and spatial domain deconvolution. Examples of the latter include, but are not limited to, differential operator methods, high-order statistical methods, maximum entropy deconvolution, and Bayesian deconvolution.

[0050] Deconvolution is often proposed as an inversion problem. Since such inversion may be unstable, smoothing techniques can be used to improve stability. Examples of such smoothing techniques include, but are not limited to, interpolation, moving average, and exponential smoothing. Another class of techniques for improving stability is regularization. Examples of regularization include, but are not limited to, Tikhonov regularization, Huber Markov regularization, discrete cosine transform regularization, and wavelet transform regularization. More details about regularization are presented below. If the inversion is stable enough, a pseudo-inverse matrix can be found. The pseudo-inverse matrix is ​​then simply multiplied by the composite array 330 to obtain the deconvolution array 340.

[0051] Figure 4 shows the process for processing the raw spectral values ​​s1...s N The first array 226 (1) and the original spectral values ​​t1 ... t M Method 400 of the second array 226 (2). Method 400 and Figure 3 The method 300 is similar in that it generates a deconvolution array 440 which is fed to Figure 1 The method 400 may also be performed by Figure 1 The signal processing circuit 130 performs the above operation.

[0052] In block 402, the first array 226(1) is deconvolved to generate deconvolved spectral values In block 406, the second array 226(2) is deconvolved into a deconvolved spectral value Each of blocks 402 and 404 is similar to block 304 of method 300. In block 406, local arrays 430(1) and 430(2) are merged to generate a deconvolved array 440. Block 406 is similar to Figure 3 Frame 302.

[0053] Figure 5 shows how fiber optic components can be used to achieve Figure 2 System 200. Figure 5, broadband light 103 emitted by light source 102 propagates along optical fiber 504. A first free space coupler 520 couples the broadband light 103 into a broadband light beam 503, which propagates through fluid 108 and exits fluid 108 as a transmitted light beam 505. A second free space coupler 522 couples the transmitted light beam 505 into input optical fiber 532. The portion of the transmitted light beam 505 coupled into optical fiber 532 is Figure 1 and Figure 2 1. The optical fiber splitter 530 divides the transmitted light 105 into a first portion 205(1) and a second portion 205(2). The optical fiber splitter 530 is Figure 2 An example of a beam splitter 230 is shown in FIG. The first portion 205(1) propagates along a first output fiber 534(1), and the second portion 205(2) propagates along a second output fiber 534(2). The first output fiber 534(1) may terminate at a first spectrometer 210(1) to pass the first portion 205(1) of the transmitted light 105 to the first spectrometer 210(1). Similarly, the second output fiber 534(2) may terminate at a second spectrometer 210(2) to pass the second portion 205(2) of the transmitted light 105 to the second spectrometer 210(2).

[0054] Fiber optic splitter 530 can be a fused bi-taper splitter, a planar lightwave circuit splitter, or another type of fiber optic splitter known in the art. Thus, fiber optic splitter 530 can be an in-line device in which optical fibers 532, 534(1), and 534(2) are connectorized pigtails. Such an in-line device with pigtails is also referred to as a fiber optic Y-cable. In other embodiments, fiber optic splitter 530 is a free space splitter that uses a free space to fiber coupler (e.g., Figure 5 Couplers 520 and 522 in the Figure 1 couple light from optical fiber 532 to optical fibers 534(1) and 534(2).

[0055] In some embodiments, Figure 2 The optical splitter 230 (eg, Figure 5 Fiber optic splitter 530 (e.g., optical fiber splitter 530) can be used to divide transmitted light 105 based on wavelength so that portions 205(1) and 205(2) have different wavelength ranges. For example, splitter 230 can be a wavelength division multiplexer. As another example, splitter 230 can use a hot mirror to reflect infrared light and transmit visible light (or use a cold mirror to transmit infrared light and reflect visible light). The reflected infrared light is used as one of portions 205(1) and 205(2), and the transmitted visible light is used as the other of portions 205(2) and 205(2).

[0056] In some embodiments, Figure 2The beam splitter 230 does not split the transmitted light 105 based on wavelength. For example, the beam splitter 230 can be a broadband beam splitter (e.g., a plate, a cube, or an optical fiber). In this case, the portions 205(1) and 205(2) have the same or similar wavelength ranges. To prevent overlap of consecutive diffraction orders in the first spectrometer 210(1), a first filter 540(1) can be used to filter the first portion 502(1) to block wavelengths outside the wavelength range of the first spectrometer. Similarly, a second filter 540(2) can be used to filter the second portion 502(2) to block wavelengths outside the wavelength range of the second spectrometer. Each of the filters 540(1) and 540(2) can be an absorption filter (e.g., a piece of colored glass) or an interference filter (e.g., a hot mirror or a cold mirror). Each of filters 540(1) and 540(2) may be a long pass filter, a band pass filter, or a short pass filter. Each of filters 540(1) and 540(2) may be another type of filter without departing from the scope thereof.

[0057] Figure 6 shows that fiber optic components can be used to achieve Figure 2 Another way of implementing the system 200. Figure 6 , a first light source 602(1) launches a first broadband light 603(1) into a first optical fiber 608(1), and a second light source 602(2) launches a second broadband light 603(2) into a second optical fiber 608(2). A fiber combiner 610 combines the broadband lights 603(1) and 603(2) into a broadband light 103, which propagates along an optical fiber 612 to a free-space coupler 520. The light sources 602(1) and 602(2), the optical fibers 608(1) and 608(2), and the fiber combiner 610 are collectively Figure 1 and Figure 2 An example of a light source 102 is shown in FIG. Fiber optic combiner 610 may be similar to fiber optic splitter 530 in that an optical combiner may be formed from optical splitters operating in reverse.

[0058] When the wavelength range of broadband light 103 needs to be wider than any single light source can generate, Figure 6. For example, the first light source 602(1) can be a broadband infrared source, and the second light source 602(2) can be a broadband visible light source. In this case, the wavelength range of the broadband light 103 spans visible light and infrared light. The broadband lights 603(1) and 603(2) can have wavelength ranges that partially overlap. Alternatively, the broadband lights 603(1) and 603(2) can have wavelength ranges that do not overlap. In one embodiment, the broadband lights 603(1) and 603(2) have respective first wavelength ranges and second wavelength ranges that are similar to the first spectrometer wavelength range and the second spectrometer wavelength range.

[0059] In other embodiments, light sources 602(1) and 602(2) emit broadband light 603(1) and 603(2) as free-space beams. In these embodiments, free-space coupler 520 and optical fibers 608(1), 608(2), and 612 are not required, and fiber combiner 610 can be replaced with a free-space beam combiner. Similarly, when fiber splitter 530 is replaced with a free-space splitter, free-space coupler 522 and optical fibers 532, 534(1), and 534(2) are not required.

[0060] Figure 7 is with Figure 2 The block diagram of system 700 is similar to system 200, except that it uses two separate light beams instead of one light beam. Specifically, system 700 includes a first light source 702(1) that emits a first broadband light beam 703(1) that propagates through fluid 108. System 700 also includes a second light source 702(2) that emits a second broadband light beam 703(2) that propagates through fluid 108. The portion of first broadband light beam 703(1) that exits test box 106 is referred to as first transmitted light beam 705(1). Similarly, the portion of second broadband light beam 703(2) that exits test box 106 is referred to as second transmitted light beam 705(2). Figure 7 , light beams 703(1) and 703(2) pass through fluid 108 as two spatially separate and distinct light beams. Thus, light beams 703(1) and 703(2) do not spatially overlap anywhere within the volume of space occupied by fluid 108.

[0061] Similar to Figure 6 , system 700 uses two light sources to measure the absorption spectrum of fluid 108 over a wider wavelength range than any single light source can generate. However, since no optical combiner (e.g., Figure 6 Fiber combiner 610), so system 700 is more Figure 6 The shown one is simpler. Since no beam splitter is required (e.g. Figure 2 The optical splitter 230 or Figure 5 Fiber splitter 530), so system 700 is also better than Figure 2 The system 200 is simpler.

[0062] Another advantage of system 700 is that it keeps light beams 703(1) and 703(2) spatially separated to prevent light contamination between them. Figure 6 , consider that the first broadband light 603(1) has a wavelength range similar to that of the first spectrometer 210(1), and the second broadband light 603(2) has a wavelength range similar to that of the second spectrometer 210(2). In this case, the fiber splitter 530 will ideally split the transmitted light 105 so that (i) the first portion 205(1) contains only light originating from the first broadband light 603(1), and (ii) the second portion 205(2) contains only light originating from the second broadband light 603(2). However, due to the imperfect performance of the fiber splitter 530, some cross-contamination light originating from the second broadband light 603(2) will be present in the first portion 205(1). As described above, such cross-contamination light may reduce the sensitivity or accuracy of the spectrometer 210(1) due to the overlap of consecutive stages. Similar arguments apply to the cross-contamination light present in the second portion 205(2). Therefore, reducing this cross-contamination light not only improves the performance of spectrometers 210(1) and 210(2), but also reduces the filtering requirements for filters 540(1) and 540(2).

[0063] When used with system 700, test box 106 may be made of two or more different materials having different light transmission properties. For example, test box 106 may include a first material through which first broadband beam 703(1) passes (but not second broadband beam 703(2)) and a second material through which second broadband beam 703(2) passes (but not first broadband beam 703(1)). The first material may be selected to achieve high transmittance at the wavelength of first broadband beam 703(1). Similarly, the second material may be selected to achieve high transmittance at the wavelength of second broadband beam 703(1). Thus, using a plurality of different materials may help increase transmittance at all wavelengths detected by spectrometers 210(1) and 210(2) compared to using a single material that may have high transmittance at some (but not all) of these wavelengths.

[0064] Although spectrometers 210(1) and 210(2) may operate simultaneously, they may alternatively operate sequentially. Figure 7In the embodiment of the present invention, the first light source 702(1) and the first spectrometer 210(1) can be operated when the second light source 702(2) does not emit the second broadband light beam 703(2) (for example, the second light source 702(2) can be turned off or blocked). Then, the first spectrometer 210(1) measures the first transmitted light beam 705(1) without measuring the cross contamination light from the second broadband light beam 703(2). After the first array 226(1) is created, the first light source 702(1) is turned off and the second light source 702(2) is turned on. Then, the second spectrometer 210(2) measures the second transmitted light beam 705(2) without measuring the cross contamination light from the first broadband light beam 703(1). This sequential operation can also be performed sequentially by the system 200. Figure 6 This is achieved by using light sources 602(1) and 602(2).

[0065] Any of the present embodiments working with two spectrometers (e.g., Figure 2 System 200, Figure 7 The system 700 of FIG. 701 and the like) can be expanded to work with three or more spectrometers. When the wavelength range exceeds the range that can be covered by two spectrometers, increasing the number of spectrometers may be useful. For example, in one embodiment, Figure 2 The system 200 also includes a third spectrometer operating in parallel with the first spectrometer 220(1) and the second spectrometer 220(2). The third spectrometer outputs a third array of raw spectral values ​​similar to the first array 226(1) and the second array 226(2). The third spectrometer may operate within a third spectrometer wavelength range that at least partially encompasses one or both of the first spectrometer wavelength range and the second spectrometer wavelength range. Alternatively, the third spectrometer wavelength range may not encompass either of the first spectrometer wavelength range and the second spectrometer wavelength range.

[0066] Signal processing circuit 130 may be adapted to process three or more arrays of raw spectral values. Similarly, methods 300 and 400 may similarly operate with three or more arrays of raw spectral values. Figure 3 In , block 302 may be expanded to merge the third array with arrays 226(1) and 226(2). Figure 4 In the example of FIG. 4 , the third original spectral value array may be deconvolved into a third local array. Block 406 may then be extended to merge the third local array with the first local array and the second local array.

[0067] Wavelength shift deconvolution

[0068] Figure 8The diagram shows how each pixel detector 116 (i) effectively combines the spectral density p(λ) with the sensitivity function f unique to the pixel detector 116 (i) during the discrete sampling of the continuous spectral density p(λ). i (λ) or kernel for convolution. The spectral density p(λ) is Figure 8 The values ​​of the spectral density p(λ) at the first wavelength λ1 are indicated by p(λ1), and the values ​​of the spectral density p(λ) at the jth wavelength are indicated by circles. For the sake of clarity, only two of these values ​​are marked. Specifically, the value of the spectral density p(λ) at the first wavelength λ1 is indicated by p(λ1), and the value of the spectral density p(λ) at the jth wavelength is indicated by p(λ1). th Wavelength λ j The value of the spectral density p(λ) at j )instruct.

[0069] Nuclear i (λ) Figure 8 is shown as having a central wavelength associated with pixel detector 116(i) The standard deviation of the Gaussian profile is denoted as σ, and thus the half-width at half maximum (HWHM) is approximately equal to 1.2σ. The HWHM is also referred to as the resolution. In an embodiment, the resolution of each pixel detector 116 (i) is between 0.2 nm and 3 nm. However, the resolution of each pixel detector 116 (i) may be greater than 0.2 nm or less than 3 nm without departing from the scope thereof. The resolution of the pixel detectors 116 (1) ... 116 (N) typically varies with the wavelength λ and is therefore not necessarily the same. The kernel f i (λ) may alternatively have a different “peak” profile, such as a Lorentzian profile, sinc or sinc 2 The middle peak value or sech of the function 2 function.

[0070] Figure 8 Also shown is a measured spectrum 806, which is simply a spectrum having N spectral values ​​m1 . . . m measured by the spectrometer 110. N (For example, Figure 1 The original spectrum values ​​s1...s N Or an array of photocurrents 118(1)...118(N)). Each spectral value m i exist Figure 8 806 is shown as a square and indicates the intensity, power or similar quantity of light. Although each pixel detector 116 (i) detects a range of wavelengths, the measured spectrum 806 is plotted so that each spectral value m i are located at the center wavelength of the corresponding pixel detector 116(i) For the sake of clarity, Figure 8Only five continuous spectral values ​​m measured by the five corresponding pixel detectors 116 (i-2), 116 (i-1), 116 (i), 116 (i+1) and 116 (i+2) are shown. i-2 、m i-1 、m i 、m i-1 and m i-2 However, the following holds for all spectral values ​​m1...m N and all pixel detectors 116 ( 1 ) . . . 116 (N) in detector array 114 .

[0071] Figure 8 shows how multiple wavelengths of light react to each spectral value m i At wavelength λ1, the core f i The value of (λ) is f i (λ1), which weights the spectral value p(λ1). Therefore, the light with wavelength λ1 has a certain effect on the spectral value m. i Contributed to item f i (λ1)p(λ1)Δλ, where Δλ is a small wavelength range over which the spectral density p(λ1) is assumed to be constant. More generally, the wavelength is λ j The light spectrum value m i Contributed to item f i (λ1)p(λ1)Δλ. Since the spectral density p(λ) and the kernel f i (λ) are both continuous functions of wavelength λ, so all wavelengths have a spectral value m i The contribution of can be expressed as the integral

[0072] m i =∫p(λ)f i (λ)dλ, (1)

[0073] where the integration is performed over all wavelengths λ.

[0074] Fig. 9 shows a wavelength λ j How does the light of the spectrum value m1...m N Several spectral values ​​contribute to . Fig. 9 Similar to Figure 8 , except that three cores are shown, specifically pixel detector 118 (i -1 ) i-1 (λ), the kernel f of the pixel detector 118(i) i (λ) and pixel detector 118 (i +1 ) i+1 (λ). For clarity, Fig. 9 Only the corresponding pixel detector 116 (i-1 ), 116(i) and 116(i +1 ) measured three continuous spectral values ​​m i-1 、m i and m i+1 However, the following holds for all spectral values ​​m1...m N And all pixel detectors 116 ( 1 ) . . . 116 (N) in detector array 114 .

[0075] At wavelength λ j At, nuclear f i The value of (λ) is f i (λ j ). For the spectral value m i The contribution of i (λ)p(λ)Δλ; This term quantifies how much light within the range Δλ will impinge on the pixel detector 118(i). However, some light within this range will also impinge on the pixel detector 118(i -1 ) and 118(i +1 ). Specifically, at wavelength λ j At, nuclear f i-1 The value of (λ) is f i-1 (λ j ). For the spectral value m i-1 The contribution of i-1 (λ)p(λ)Δλ. Similarly, at wavelength λ j At, nuclear f i+1 The value of (λ) is f i+1 (λ), resulting in a spectral value m i+1 The contribution of f i+1 (λ)p(λ)Δλ.

[0076] although Fig. 9 Not shown, but the wavelength is λ j The light of m will also impinge on other pixel detectors 116, thereby affecting other spectral values ​​(e.g., m i±2 、m i±3 The wavelength reaching the pixel detector 118(k) is λ j The amount of light is determined by the kernel f of pixel detector 118(k). k (λ) is given, and thus with the wavelength difference The square of decreases exponentially. When it is large enough, the value f k (λ j ) will be very small so that the contribution f k (λ j )p(λ j )Δλ is negligible and can therefore be neglected. Therefore, for the spectral value mi The wavelength that contributes the most is usually the one closest to the center wavelength Those wavelengths (for example, at the center wavelength within one analytical resolution or standard deviation σ of .

[0077] Figure 8 and Fig. 9 Also shown is a diagram for converting the spectral values ​​m1 ...m N Deconvolution is a method for estimating the spectral density p(λ), which is greater than the spectral values ​​m1...m N This deconvolution can be done by Figure 1 The signal processing circuit 130 (for example, by converting the original spectrum values ​​s1 . . . s N The array 126 is deconvolved into the deconvolved spectral values Using this method, Formula 1 can be approximated as:

[0078]

[0079] in are R reference wavelengths, and Δ1...Δ R There are R wavelength ranges. Each wavelength range Δ k ∈{Δ1...Δ R} include the corresponding reference wavelength Wavelength range Δ1...Δ R can be different values ​​of R. Alternatively, when the reference wavelength When uniformly distributed, the wavelength range is Δ1...Δ R Can have a common value.

[0080] To extend Equation 2 to all N pixel detectors 116 in detector array 114, we use the following vector notation:

[0081]

[0082] In Equation 3, F is a kernel matrix whose rows correspond to the N pixel detectors 116 and whose columns correspond to the R reference wavelengths Each element F of the kernel matrix F ij All given values The expansion of Equation 2 can then be expressed as

[0083] m=Fp. (4)

[0084] The goal is to find a solution based on m and F If the kernel matrix F is well-conditioned, the pseudo-inverse F of the matrix F can be calculated + . The solution is then given by

[0085]

[0086] If F is one-to-one, then

[0087] F + =(F T F) -1 F T . (6)

[0088] However, when F is ill-conditioned, the pseudo-inverse F + It is likely that it cannot be calculated. In this case, the solution Another technique is to use least squares regression to solve for

[0089]

[0090] Change p.

[0091] When F is ill-conditioned, least squares regression may still not give the correct solution. In such cases, one or more types of regularization can be used. One such regularization is Tikhonov regularization. In this case, by adding the regularization term to modify Equation 7, where Γ is the Tikhonov matrix. The resulting Tikhonov regularized least squares regression is given by

[0092]

[0093] There are many choices for the Tikhonov matrix Γ known in the art, any of which can be used. One choice is Γ = αI, where and I is the identity matrix of size R. In this case, the regularization term penalizes large elements in p. Larger values ​​of α produce stronger regularization, while smaller values ​​of α produce weaker regularization, with α = 0 producing the unregularized least squares regression of Equation 7. Another choice for the Tikhonov matrix Γ is Γ = I - α k I k , where α k is a constant, and I k is along the kth th The offset matrix has nonzero elements on either the upper or lower diagonal. This choice of Γ creates a finite difference matrix that penalizes p by k th In one embodiment, only the first and second upper diagonals are non-zero (i.e., for all k ≥ 3, α k =0).

[0094] The above examples of Γ can be added together to obtain the generalized Tikhonov matrix Tikhonov regularization produces a closed-form solution, just like unregularized least squares regression. The solution is given by

[0095]

[0096] When F is one-to-one,

[0097] F + =(F T F+Γ T Γ) -1 F T . (10)

[0098] In an embodiment, the central wavelength Form an ordered sequence and reference wavelength It should be noted that the central wavelength is not required to Any of the two is equal to the reference wavelength In contrast, many prior art deconvolution techniques only consider R = N for all i and Thus, the deconvolution described herein not only “sharpenes” the characteristics of the measured spectrum 806, much like prior art deconvolution techniques, but also allows the deconvolution to be performed. Each component Different from the central wavelength This type of deconvolution (where the reference wavelength At least one of them is different from all central wavelengths ) is referred to in this article as wavelength-shift deconvolution.

[0099] Wavelength-shift deconvolution advantageously accounts for the fact that different spectrometers, even of the same type, may have central wavelengths These differences are typically caused by manufacturing variations (e.g., component manufacturing and assembly tolerances). However, the machine learning model 140 will use only the wavelengths that are consistent with the reference wavelength. The reference wavelengths may be different from the central wavelength of the current spectrometer. Wavelength-shifted deconvolution accounts for these differences, thereby ensuring that each deconvolved spectral value fed into the machine learning model 140 Both the machine learning model 140 considers the deconvolution spectral value Associated reference wavelength This wavelength shift improves the accuracy of the indication 142 output by the machine learning model 140.

[0100] Since the reference wavelength The number R may be related to the central wavelength The number N of pixel detectors is different, so wavelength shift deconvolution also allows the machine learning model 140 to be used with spectrometers with different numbers of pixel detectors. Specifically, wavelength shift deconvolution can be used to transform N original spectral values ​​into R deconvolved spectral values. For example, consider a first spectrometer with a first detector array having N1=2048 pixel detectors, a second spectrometer with a second detector array having N2=3648 pixel detectors, and a machine learning model with R=2500 inputs (e.g., input nodes of an artificial neural network). Here, a first kernel matrix F1 can be constructed to transform the N1=2048 spectral values ​​m1...m 2048 Transformed into R = 2500 wavelength shifted and deconvoluted spectral values They are fed into the machine learning model 140. Similarly, a second kernel matrix F2 can be constructed to convert another N2 = 3648 spectral values ​​m1 ... m measured with the second detector array into 3648 Transformed into R = 2500 wavelength shifted and deconvoluted spectral values They are fed into the machine learning model 140. In both cases, the same machine learning model 140 is used. Although this example shows that R can be greater or less than N, R is equal to N instead. Furthermore, when R is equal to N, the reference wavelength Each of the The corresponding central wavelength in (i.e., In contrast, wavelength-shift deconvolution can be used without changing the number of spectral values ​​(in this case, the kernel matrix F will be a square matrix).

[0101] Wavelength shift deconvolution also helps manufacturing and technical support by facilitating modularity of any of the systems described herein. For example, during production, all spectrometers can be tested using the same machine learning model 140. For example, consider a system in the field where a spectrometer has been damaged. The damaged spectrometer can be easily replaced with one that may have a different wavelength calibration (i.e., a different center wavelength). ) of a new spectrometer. The spectrometer may include a memory that stores its own central wavelength When a new spectrometer is first connected to the system, the signal processing circuit 130 may retrieve the central wavelength from the memory. The signal processing circuit 130 can then use these center wavelengths to construct a new kernel matrix F. Alternatively, the kernel matrix F may be stored in the memory of the spectrometer, in which case the signal processing circuit 130 may retrieve the kernel matrix F from the memory of the spectrometer without performing any matrix construction. In any case, the machine learning model 140 does not need to be modified at all.

[0102] Using the same machine learning model 140 for all units is also advantageous because training data can be expensive and difficult to obtain. Therefore, it is easier, faster, and cheaper to maintain a single "master" model that is updated with new training data and then pushed to all deployed units (e.g., via the Internet). Alternatively, when the machine learning model 140 is deployed on a remote computer server that communicates with the field-deployed units, the internal architecture of the machine learning model 140 (i.e., the architecture other than multiple inputs and multiple outputs) can be modified without changing the way the field-deployed units communicate with the remote computer server.

[0103] Construct the kernel matrix F

[0104] To construct the kernel matrix F of a given spectrometer, we first need to know the central wavelength and reference wavelength The reference wavelength may be selected based on characteristics of one or more components in the fluid 108. Therefore, some embodiments of the present method include selecting a reference wavelength steps.

[0105] The central wavelength can be easily measured using current spectrometers For center wavelength Any determination of is referred to as wavelength calibration. As an example of wavelength calibration, consider an adjustable monochromatic light source (e.g., a tunable laser or a broadband light source filtered by a monochromator) whose wavelength can be precisely and accurately measured (e.g., with a wavelength meter). When this monochromatic light is coupled into the spectrometer 110, the photocurrent 118(i) output by the pixel detector 116(i) can be measured at several different wavelengths at which the photocurrent 118(i) is maximum. As described above, these measurements will form a Gaussian profile (or a similar "peaked" distribution). The measurements can be fit (e.g., nonlinear regression) to a best fit function to obtain best fit values ​​for the center wavelength, width (or analytical resolution), and amplitude of the Gaussian profile. The best fit function estimates the sensitivity function f of the pixel detector 116(i) i This process can be repeated to obtain the sensitivity function f for all N pixel detectors 116(1)...116(N) i The estimated value of (λ).

[0106] Since the sensitivity function f of all pixel detectors 116 is measured i(λ) can be time consuming, so certain assumptions can be made to speed up the process. For example, the center wavelength Some of them can be determined by interpolation. For example, when measuring the sensitivity function f1(λ) of the first pixel detector 116(1) to obtain the first central wavelength And measure the sensitivity function f of the last pixel detector 116 (N) N (λ) to obtain the last central wavelength All intermediate central wavelengths can be estimated by linear interpolation. These intermediate center wavelengths are assumed to be uniformly distributed. The resolution of the pixel detectors 116 (i.e., the Gaussian width) can be similarly obtained by linearly interpolating the best-fit resolutions of the first pixel detector 116 (1) and the last pixel detector 116 (N). Alternatively, it can be simply assumed that all pixel detectors 116 have the same resolution.

[0107] Many spectrometer manufacturers perform wavelength calibration before shipment. Available to customers or users, they can be stored in a memory physically co-located with the spectrometer. Alternatively, the center wavelength It may be printed on paper shipped with the goods, provided electronically (e.g., on a memory stick), or may be downloaded from a remote server. In any case, when the manufacturer (or another third party) performs the wavelength calibration and provides the resulting center wavelength , the user typically does not need to repeat the process. Thus, some embodiments of the present method can be used to construct the kernel matrix F without performing any type of calibration measurements on the spectrometer. In this case, the sensitivity function f of the pixel detector 116 (i) can be considered to be i (λ) has a central wavelength The Gaussian width or resolution of pixel detector 116 (i) may similarly be considered to have a value provided by the manufacturer (e.g., on a specification sheet or test data sheet) or interpolated therefrom.

[0108] The kernel matrix F has N×R elements, which are labeled F ij , where i indexes the pixel detectors 116 (i.e., i ranges from 1 to the number n of pixel detectors 116), and j indexes the reference wavelength (i.e., j ranges from 1 to the reference wavelength The number of R). Elements F ij The value of Therefore, once the sensitivity functions of all pixel detectors 116 are known, the sensitivity functions can be calculated by evaluating the R reference wavelengths Each sensitivity function f at each ofi To fill the kernel matrix F.

[0109] Fig.10 is a functional diagram of a computing system 1000, which is Figure 1 The computing system 1000 can be implemented, for example, with other components of any system implementation described herein (e.g., Figure 1 System 100, Figure 2 The computing system 1000 includes a read-only memory (ROM) 1006 and a random access memory (RAM) 1008, which communicate with the processor 1004 via the system bus 1002. In some embodiments, the computing system 1000 also includes a graphics display 1014 for intuitively displaying information to a user, receiving input from a user, or both. Alternatively, the computing system 1000 may include a display adapter for use with a graphics display provided by a third party.

[0110] The computing system 1000 also includes an input / output (I / O) block 1010 that interfaces with one or more spectrometers to receive arrays of raw spectral values. For example, the I / O block 1010 may interface with Figure 2 The spectrometers 210(1) and 210(2) are connected to receive the raw spectral values ​​s1...s N The first array 226 (1) and the original spectral values ​​t1 ... t M The computing system 1000 also includes an I / O block 1012 through which the computing system can communicate with peripheral devices or remote computer systems (e.g., a hard drive, a USB port, a memory card, a network connector, etc.). The I / O blocks 1010 and 1012 are also connected to the system bus 1002 and can therefore communicate with the processor 1004 and store data in and retrieve data from the RAM 1008.

[0111] Processor 1004 can be any type of circuit capable of performing logic, control and input / output operations. For example, processor 1004 may include one or more of a microprocessor, a graphics processing unit (GPU), a digital signal processor (DSP), a field programmable gate array (FPGA), a system on chip (SoC) and a microcontroller unit (MCU) with one or more central processing unit (CPU) cores. Processor 1004 may also include a memory controller, a bus controller, one or more coprocessors and / or other components of the management processor 1004 and the data stream between other components that can be communicatively coupled to the system bus 1002. Processor 1004 may be implemented as a single integrated circuit (IC) or multiple ICs. In some embodiments, one or more of processor 1004, ROM 1006, RAM 1008, I / O block 1010 and I / O block 1012 is implemented as a single IC. Processor 1004 may adopt a complex instruction set computing (CISC) architecture, or a reduced instruction set computing (RISC) architecture.

[0112] ROM 1006 stores machine-readable instructions (in Fig.10 These machine-readable instructions, shown as firmware 1020 in the figure, control the computing system 1000 when executed by the processor 1004 to implement the functions described herein. Fig.10 , firmware 1020 includes a kernel matrix generator 1022 that constructs and populates the kernel matrix F. Firmware 1020 also includes an array combiner 1024 that combines arrays 226(1) and 226(2) into a composite array 330, thereby implementing block 302 of method 300. Firmware 1020 also includes a deconvolver 1026 that deconvolves the composite array 330 to generate a deconvolved array 340, thereby implementing block 304 of method 300. Firmware 1020 also includes an indication generator 1028 that feeds the deconvolved array 340 into the machine learning machine 140 to obtain indications 142. Firmware 1020 also includes an outputter 1030 that outputs one or more of indication 142 (e.g., by displaying indication 142 on graphics display 1014 or transmitting indication 142 to an external computer via I / O block 1012), first array 1042(1), second array 1042(2), composite array 1044, deconvolution array 1046 (e.g., by drawing deconvolution array 1046 on graphics display 1014), and other data 1040. ROM 1006 may store additional machine-readable instructions (e.g., operating system instructions, I / O control instructions, etc.) or data without departing from the scope thereof. ROM 1006 may be implemented using non-volatile memory such as flash memory, NVRAM, FRAM, MRAM, EEPROM, EPROM, or any combination thereof.

[0113] RAM 1008 stores data 1040 used by processor 1004 when executing firmware 1020. Fig.10 , data 1040 includes a first raw spectral value array 1042(1) (e.g., Figure 2 The original spectral values ​​s1...s N 226 (1)), the second raw spectral value array 1042 (2) (e.g., Figure 2 The original spectral values ​​t1...t M of the second array 226(2)), the composite array 1044 (e.g., Figure 3 The original spectral values ​​c1...c N+M composite array 330), deconvolution array 1046 (e.g., Figure 3 The deconvoluted spectrum value in The deconvolution array 340), the kernel matrix F (eg, see formula 3), the center wavelength Reference wavelength Instruction 1048 (e.g., Figure 1 142) and a trained machine learning model 1050 (e.g., Figure 1 However, the RAM 1008 may store additional data 1040 without departing from the scope thereof.

[0114] RAM 1008 may be implemented using volatile memory such as DRAM, SRAM, or a combination thereof. Reference wavelength and machine learning models 1050 are not expected to change frequently, so they may instead be stored in non-volatile memory (e.g., ROM 1006), thereby ensuring that the computing system 1000 retains this information during power outages.

[0115] Each of the I / O blocks 1010 and 1012 can implement a communication protocol. For example, each of the I / O blocks 1010 and 1012 can be a serial communication interface (e.g., RS-232, RS-422, RS-485, etc.), a parallel communication interface (e.g., GPIB, PCI, SCSI, etc.), a synchronous serial communication interface (e.g., I2C, SPI, SSC, etc.), a universal serial bus (USB) interface, a multimedia card interface (e.g., SD card, compact flash, etc.), a wired network interface (e.g., Ethernet, Infiniband, Fiber Channel, etc.), a wireless network interface (e.g., WiFi, Bluetooth, BLE, ZigBee, ANT, etc.), a cellular network interface (e.g., 3G, 4G, 5G, LTE), an optical network interface (e.g., SONET, SDH, IrDA, etc.), and a field bus interface. In some embodiments, the computing system 1000 uses a separate I / O block for each spectrometer. Thus, in these embodiments, computing system 1000 may include additional I / O blocks such that each of spectrometers 210(1) and 210(2) communicates with computing system 1000 via dedicated I / O blocks.

[0116] although Fig.10 1000, but the computing system 1000 may include additional I / O functionality as needed for communicating with additional peripheral devices. For example, the computing system 1000 may include one or more additional network interfaces (e.g., an Ethernet port or a WiFi adapter) to receive firmware 1020 from ROM 1006, update firmware 1020, update machine learning model 1050, or feed deconvolution array 1046 to a machine learning model executed on a remote computer server. Similarly, one or more additional network interfaces may be used to obtain one or more of the following: center wavelength Reference wavelength and machine learning model 1050.

[0117] The kernel matrix F may be stored in RAM 1008 (or ROM 1006) as a two-dimensional array. Each cell or element F of the kernel matrix F ij The first index value i uniquely identifies one of the pixel detectors 216 and 218. The second index value j uniquely identifies the reference wavelength. Unit F ij Stores the spectral response of the pixel detector 116(i) identified by the first index value i The spectral response At the reference wavelength identified by the second index value j Assessment will be conducted at the relevant department.

[0118] Deconvolver 1026 may implement the wavelength shift deconvolution described above, or another deconvolution technique known in the art. Deconvolver 1026 may implement least squares regression with or without regularization. When least squares regression is implemented using Tikhonov regularization, RAM 1008 or ROM 1006 may additionally store the Tikhonov matrix Γ (e.g., see Equation 8).

[0119] In some embodiments, the processor 1004 does not execute machine-readable instructions (e.g., FPGA) to implement the functions described herein. Instead, the processor c1004 is pre-programmed to perform tasks and therefore acts as a hard-wired circuit. Therefore, in these embodiments, the functions are implemented only in hardware and firmware 1020 may not be included. In other embodiments, the functions are implemented only in software. In other embodiments, the functions are implemented as a combination of hardware and software.

[0120] Although Fig.10 The computing system 1000 is shown as having a system bus 1002, but the computing system 1000 may be implemented with different types of architectures without departing from the scope of the present invention. For example, when the firmware 1020 and the data 1040 are stored in separate memories, such as Fig.10 As shown, the firmware 1020 and the data 1040 may be transferred to / from the processor 1004 using separate buses. In this case, the firmware 1020 and the data 1040 may be stored in separate memory spaces, thereby implementing the Harvard architecture. Alternatively, the processor 1004 may include one or more layers of cache so that the computing system 1000 uses only one system bus 1002 to implement a modified Harvard architecture. In some embodiments, the firmware 1020 is stored as an application in a secondary storage device (e.g., a hard disk) and is loaded into the RAM 1008 at power-on (i.e., boot-up). In this case, the application and the data 1040 share the same memory space, thereby implementing the von Neumann architecture.

[0121] Other Implementations

[0122] Medical centers (e.g., doctor's offices, urgent care centers, pharmacies, etc.) often have long wait times and high costs due to their reliance on highly skilled and highly paid medical professionals. These professionals, while efficient and mostly required by law, create multiple bottlenecks. The present embodiment includes a ProSpectral system of pattern computers that can reduce the number and severity of these bottlenecks by automating a portion of medical services. Theoretically, the ProSpectral system can diagnose a variety of diseases using as little as two drops of human fluid (e.g., saliva). Due to its speed and ease of use, several versions of the ProSpectral system can be used as self-service diagnostic and pharmacy stations. The station (which can be constructed as a consulting room) allows people who believe they are unwell to receive diagnosis and treatment without waiting for a face-to-face visit with a medical professional. Using a self-service station also reduces the chance of spreading infectious diseases to other medical staff and patients.

[0123] As an example of how the ProSpectral system can be used as a self-service kiosk, consider someone who thinks they may have the flu. Under prior art methods for providing medical care routinely, the patient may go to an urgent care facility for a diagnosis and, if positive, receive antiviral treatment. The patient may need to wait (e.g., 30 minutes to 60 minutes or more) for a visit, at which point the medical professional needs to perform a nasal swab test, wait 10 minutes to 20 minutes for the results, and then prescribe and fill the medication or give the patient an injection. The entire process can easily last an hour, or even longer.

[0124] In an alternative scenario, after arriving at an urgent care facility, the patient will wait for an available ProSpectral office. Once in the office, the person will identify themselves. For example, the patient may present identification to be scanned at the office (e.g., government ID, insurance card, etc.) or log into the system using predetermined credentials (e.g., username and password). The person will place two drops of saliva into a cuvette. For example, this task can be accomplished by dispensing a cuvette into a container. As another example, a pipette is dispensed to fill the cuvette, which is secured by the office and exposed through a small opening. The sample will then be scanned. A diagnosis may then be presented or the patient may be referred to a medical professional (e.g., if the results are inconclusive). If the patient is diagnosed (e.g., has the flu), the office may provide a prescription (e.g., an antiviral medication). This may legally require the intervention of a medical professional, which may be performed remotely via a video screen, camera, speaker, and microphone integrated into the office. At this point, the person leaves the office and the medical center. The office may automatically lock and self-clean by disinfecting spray and / or exchanging the air in the office. At this point, the clinic is disinfected and ready to receive another person.

[0125] To reduce waiting time, multiple clinics can be installed at a medical facility. To reduce the cost of more clinics, the ProSpectral system can be configured with multiple clinics that communicate with a single sensor (e.g., Figure 1 System 100, Figure 2 System 200, Figure 7 Each clinic can also communicate with the sensor (e.g., wirelessly or by wire) to send and receive digital information (e.g., Figure 1 Instructions 142).

[0126] The ProSpectral system can also be used in medical research and clinical trials. Examples include scientific discovery, treatment monitoring, screening for specific diseases, dose response of patient medications, adverse reaction screening, and epidemiological population monitoring. Many of these uses are conducive to improving the prediction of drug efficacy, efficiency, sensitivity, toxicity, and side effects (e.g., accuracy, timeliness, etc.). Many of these uses also improve the identification of patients who are or are not suitable for participating in trials and / or final treatment. Specifically, the ProSpectral system can also be used to promote early intervention, that is, to identify early warning signs of possible adverse reactions, especially for drugs with a high percentage of side effects (e.g., 3% to 5%).

[0127] Many prior art instruments and techniques are used to plan, conduct, and improve clinical trials. These can include any number of tests to assess the impact of treatments and the health of subjects. The nature of the tests varies widely (e.g., physical\blood pressure, chemical reagents\blood tests, cultures\bacterial infections, DNA sequencing\genetic conditions). Most biomedical tests must use well-established sample types (e.g., blood, urine, tissue) and biochemistries (e.g., reagents). Most subsequent bioinformatics / biostatistical analyses rely on well-known mathematics that can only verify hypotheses about patterns and correlations, but cannot intrinsically discover and explain them.

[0128] The present embodiment helps to overcome many of these limitations. An example of such an improvement is the ability to replace multiple individual tests with one multi-purpose multiplexed test. In general, the disadvantages of being limited to using individual tests include their invasiveness, time to results, cost, single use, consumption of consumables and reagents, infrastructure requirements, and required medical clearances. By helping to overcome these disadvantages, the present embodiment can reduce the time and cost of clinical trials.

[0129] Many of the embodiments in this embodiment have several advantageous properties, such as:

[0130] ·Results in three seconds significantly speed up diagnosis, improve patient turnover and satisfaction;

[0131] Balanced accuracy of 98.5% matches the accuracy of PCR tests;

[0132] No reagents required – the facility does not need to handle or dispose of toxic waste;

[0133] Only two drops of saliva are required—no invasive, unpleasant nasopharyngeal swab required;

[0134] Small and portable (shoebox size and weighs less than 3.5kg)—suitable for mobile testing;

[0135] ·Can process more than 240 samples per hour;

[0136] Simple operation, training and maintenance—no medical approval required;

[0137] Detection of new diseases and variants requires only remote software updates;

[0138] Cheaper than other tests that require reagents and complex equipment.

[0139] As implemented in many embodiments of this embodiment, the pattern discovery engine (PDE) of the pattern computer can find patterns in biomarker absorption features that are reliably associated with a specific disease infection and create analytical models to test for the disease. It can also identify patterns in the data that are ultimately determined to be features of other events and states (e.g., toxicity, side effects).

[0140] The PDE processes the data that ProSpectral reads from the saliva sample as well as external data (usually obtained only in response to a question) about the subject's health status (e.g., infection with a disease, indicators of toxicity, indicators of drug effectiveness, co-morbidities, susceptibility to allergies / side effects, etc.). This allows it to recognize the patterns it finds in the spectrophotometric data as signatures of these other factors. Once the association is established, ProSpectral can report results for those factors independently and much earlier than is typically realized. This improves the safety, planning, execution, timelines, and costs of clinical trials and studies. Once a baseline saliva scan of a subject is collected before the trial / study begins, longitudinal testing (repeated testing during the study) enables the creation of models that can:

[0141] Identify which candidate subjects are suitable or not suitable for trial participation;

[0142] Monitor changes in saliva scan data that correlate with current or future changes in effectiveness / toxicity / side effects and proactively adjust (e.g., dosage);

[0143] Predict which subjects may be more susceptible to toxicity or side effects;

[0144] Predicting treatment efficacy in individual subjects; and

[0145] Identify changes that predict problems that usually only occur after later / higher doses.

[0146] The technology used by the ProSpectral system is fundamentally different from most current related over-the-counter disease detection technologies. It detects host metabolic responses rather than the disease itself (e.g., a virus). Different disease states leave specific traces (signatures) in the molecular composition of these biomarkers. Although biological samples have the same chemical groups, the bond structures and chemical abundances can vary greatly. The ProSpectral system uses fine spectral resolution and accurate detection of chemical bond lengths and angles (respectively) to accurately identify the presence of a specific disease state. and 1°) sensitive signals to analyze them accurately.

[0147] Importantly, it is a multiplexed device by design that is intended to detect additional diseases and variants and is capable of analyzing other body fluids (i.e., blood and urine) as the need evolves. Updating the device to detect additional diseases or new variants requires only a software update that can be performed remotely.

[0148] Feature combination

[0149] The above-mentioned characteristics and those characteristics claimed below can be combined in various ways without departing from the scope of the present invention. The following examples illustrate possible non-limiting combinations of the above-mentioned characteristics and embodiments. It should be clear that other changes and modifications can be made to this embodiment without departing from the spirit and scope of the present invention:

[0150] (A1) A system includes: a light source operable to emit broadband light; and a cartridge holder shaped to receive a test cartridge confining a fluid. The cartridge holder is positioned so that the broadband light passes through the test cartridge when emitted by the light source. The system also includes a spectrometer that, in response to receiving the broadband light after the broadband light passes through the test cartridge, disperses the broadband light onto a plurality of pixel detectors of a detector array. The system also includes a digitizer electrically connected to the detector array and operable to read a raw spectral value array from the detector array. The system also includes a signal processing circuit that is programmed to deconvolve the raw spectral value array based on the spectral response of each of the plurality of pixel detectors to generate a deconvolved spectral value array. The signal processing circuit is also programmed to feed the deconvolved spectral value array into a trained machine learning model that processes the deconvolved spectral value array to obtain an indication of the presence of one or more components in the fluid. The signal processing circuit is also programmed to output the indication.

[0151] (A2) In the system shown in (A1), the spectrometer operates within a certain wavelength range and the broadband light spans the wavelength range.

[0152] (A3) In the system denoted as (A2), the wavelength range includes at least a portion of the visible region of the electromagnetic spectrum.

[0153] (A4) In any of the systems shown in (A2) and (A3), the wavelength range is 350 nm to 750 nm.

[0154] (A5) In any of the systems denoted as (A2) to (A4), the wavelength range includes at least a portion of the near infrared region of the electromagnetic spectrum.

[0155] (A6) In any of the systems shown in (A2) to (A5), the wavelength range is 900 nm to 1700 nm.

[0156] (A7) In any one of the systems shown in (A1) to (A6), the resolution of the spectrometer is in the range of 0.2 nm to 3 nm (inclusive).

[0157] (A8) In any of the systems denoted as (A1) to (A7), the spectrometer is a spectrophotometer.

[0158] (A9) In the system described in (A1) to (A8), the cartridge holder includes a cuvette holder, and the test cartridge includes a cuvette.

[0159] (A10) In the system shown in (A9), the cuvette comprises glass, plastic, or a combination thereof.

[0160] (A11) In any of the systems denoted as (A1) to (A10), the fluid comprises human saliva.

[0161] (A12) In any of the systems denoted as (A1) to (A11), the fluid comprises a biological fluid obtained from a human or an animal.

[0162] (A13) In any of the systems described in (A1) to (A12), the signal processing circuit includes a memory to store a kernel matrix having a plurality of cells. Each cell of the plurality of cells is indexed by a first index value of a plurality of first index values ​​and a second index value of a plurality of second index values. The first index value uniquely identifies one of the plurality of pixel detectors. The second index value uniquely identifies one of a plurality of reference wavelengths. Each cell stores the spectral response of the one of the plurality of pixel detectors identified by the first index value of the each cell, the spectral response being located at the one of the plurality of reference wavelengths identified by the second index value of the each cell. The signal processing circuit is also programmed to deconvolve the raw spectral value array by performing a least squares regression based at least on the raw spectral value array and the kernel matrix.

[0163] (A14) In the system shown in (A13), the signal processing circuit is programmed to perform least squares regression with Tikhonov regularization.

[0164] (A15) In any of the systems denoted as (A1) to (A14), the signal processing circuit includes a field programmable gate array.

[0165] (A16) In any of the systems denoted as (A1) to (A15), the signal processing circuit includes a processor and a memory communicatively coupled to the processor. The signal processing circuit further includes a deconvolution engine implemented as machine-readable instructions stored in the memory and, when executed by the processor, controlling the signal processing circuit to deconvolve the original spectral value array to obtain the deconvolved spectral value array.

[0166] (A17) In the system shown in (A16), the memory stores a kernel matrix including multiple cells. Each cell of the multiple cells is indexed by a first index value of a plurality of first index values ​​and a second index value of a plurality of second index values. The first index value uniquely identifies one of the multiple pixel detectors. The second index value uniquely identifies one of a plurality of reference wavelengths. Each cell stores the spectral response of the one of the multiple pixel detectors identified by the first index value of each cell, and the spectral response is located at the one of the multiple reference wavelengths identified by the second index value of each cell. The machine-readable instructions that control the signal processing circuit to perform deconvolution when executed by the processor include machine-readable instructions that control the signal processing circuit to perform least squares regression based on at least the original spectral value array and the kernel matrix when executed by the processor.

[0167] (A18) In the system shown in (A17), the machine-readable instructions that control the signal processing circuit to perform least squares regression when executed by the processor include machine-readable instructions that control the signal processing circuit to perform least squares regression with Tikhonov regularization when executed by the processor.

[0168] (A19) In any of the systems denoted as (A1) to (A18), the signal processing circuit is further programmed to feed the deconvolved spectral value array into the trained machine learning model by transmitting the deconvolved spectral value array to an external computer system, the external computer system feeding the deconvolved spectral value array into the trained machine learning model to obtain the indication. The signal processing circuit is further programmed to receive the indication from the external computer system.

[0169] (A20) In any of the systems denoted as (A1) to (A19), the test kit does not contain any reagents.

[0170] (B1) A method comprising: transmitting broadband light through a fluid confined within a test box; dispersing the broadband light onto a plurality of pixel detectors forming a detector array after the broadband light passes through the test box; reading a raw spectral value array from the detector array; deconvolving the raw spectral value array based on the spectral response of each of the plurality of pixel detectors to generate a deconvolved spectral value array; feeding the deconvolved spectral value array into a trained machine learning model, the machine learning model processing the deconvolved spectral value array to obtain an indication of the presence of one or more components in the fluid; and outputting the indication.

[0171] (B2) In the method shown in (B1), the method also includes collecting the fluid.

[0172] (B3) In the method shown in (B2), the collecting does not include adding any reagent to the fluid.

[0173] (B4) In any of the methods denoted as (B2) and (B3), the collecting comprises injecting at least a portion of the fluid into the test cartridge and placing the test cartridge in the cartridge holder.

[0174] (B5) In any of the methods shown as (B2) to (B4), the collecting includes collecting the fluid from a human patient.

[0175] (B6) In the method shown in (B5), the collecting includes collecting saliva from the human patient.

[0176] (B7) In any of the methods shown in (B5) and (B6), the method also includes diagnosing the human patient with a disease based on the indication.

[0177] (B8) In the method shown in (B7), the method also includes providing therapeutic intervention for treating the disease to the human patient.

[0178] (B9) In the method shown in (B8), the therapeutic intervention is a surgical procedure, a non-surgical medical procedure, a prescription of one or more medications, or a combination thereof.

[0179] (B10) In any of the methods shown in (B1) to (B9), the deconvolution includes performing least squares regression based on at least the original spectral value array and a kernel matrix comprising multiple cells; each of the multiple cells is indexed by a first index value among a plurality of first index values ​​and a second index value among a plurality of second index values; the first index value uniquely identifies one of the multiple pixel detectors; the second index value uniquely identifies one of a plurality of reference wavelengths; and each cell stores the spectral response of the one of the multiple pixel detectors identified by the first index value of each cell, and the spectral response is located at the one of the multiple reference wavelengths identified by the second index value of each cell.

[0180] (B11) In the method shown in (B10), performing least squares regression includes performing least squares regression with Tikhonov regularization.

[0181] (B12) In any of the methods shown in (B1) to (B11), the feeding includes transmitting the deconvolved spectral value array to an external computer system, wherein the external computer system feeds the deconvolved spectral value array into the trained machine learning model. The feeding also includes receiving the indication from the external computer system.

[0182] (C1) A system includes: a light source operable to emit broadband light; and a cartridge holder shaped to receive a test cartridge confining a fluid. The cartridge holder is positioned so that the broadband light passes through the test cartridge when emitted by the light source. The system also includes a first spectrometer that, in response to receiving a first portion of the broadband light after the broadband light has passed through the test cartridge, disperses a first wavelength range of the first portion of the broadband light onto a first plurality of pixel detectors forming a first detector array. The system also includes a second spectrometer that, in response to receiving a second portion of the broadband light after the broadband light has passed through the test cartridge, disperses a second wavelength range of the second portion of the broadband light onto a second plurality of pixel detectors forming a second detector array. The system also includes a digitizer that is electrically connected to the first detector array and the second detector array. The digitizer is operable to read a first array of raw spectral values ​​from the first detector array and a second array of raw spectral values ​​from the second detector array. The system also includes a signal processing circuit that is programmed to deconvolute the first raw spectral value array and the second raw spectral value array to generate a deconvolved spectral value array based on the spectral response of each of the first plurality of pixel detectors and the second plurality of pixel detectors; feed the deconvolved spectral value array into a trained machine learning model, which processes the deconvolved spectral value array to obtain an indication of the presence of one or more components in the fluid; and output the indication.

[0183] (C2) In the system shown in (C1), the signal processing circuit is also programmed to (i) merge the first raw spectral value array and the second raw spectral value array into a raw spectral value composite array, and (ii) deconvolute the raw spectral value composite array to obtain the deconvolved spectral value array.

[0184] (C3) In the system shown in (C1), the signal processing circuit is also programmed to deconvolve the first original spectral value array to generate a first deconvolved spectral value local array; deconvolve the second original spectral value array to generate a second deconvolved spectral value local array; and merge the first deconvolved spectral value local array and the second deconvolved spectral value local array to obtain the deconvolved spectral value array.

[0185] (C4) In any of the systems denoted as (C1) to (C3), the first wavelength range includes at least a portion of the visible region of the electromagnetic spectrum, and the second wavelength range includes at least a portion of the near infrared region of the electromagnetic spectrum.

[0186] (C5) In the system shown in (C4), the first wavelength range includes 350nm to 700nm, and the second wavelength range includes 900nm to 1700nm.

[0187] (C6) In any of the systems denoted as (C1) to (C5), the first wavelength range and the second wavelength range partially overlap.

[0188] (C7) In any of the systems denoted as (C1) to (C5), the first wavelength range and the second wavelength range do not overlap.

[0189] (C8) In any of the systems denoted as (C1) to (C7), the first spectrometer has a first resolution, and the second spectrometer has a second resolution that is different from the first resolution.

[0190] (C9) In any of the systems denoted as (C1) to (C8), each of the first spectrometer and the second spectrometer comprises a spectrophotometer.

[0191] (C10) In any of the systems described in (C1) to (C9), the system further includes: an input fiber optic cable; a first spectrometer fiber optic cable and a second spectrometer fiber optic cable; a fiber collimator that couples the broadband light into the input fiber optic cable after the broadband light passes through the test box; and a beam splitter that splits the broadband light leaving the input fiber optic cable into a first portion and a second portion, couples the first portion into the first spectrometer fiber optic cable, and couples the second portion into the second spectrometer fiber optic cable. The first spectrometer fiber optic cable couples the first portion into the first spectrometer. The second spectrometer fiber optic cable couples the second portion into the second spectrometer.

[0192] (C11) In any of the systems denoted as (C1) to (C10), the cartridge holder comprises a cuvette holder that positions a cuvette such that the fluid is illuminated by the broadband light while confined within the cuvette.

[0193] (C12) In the system shown in (C11), the cuvette comprises glass, plastic, or a combination thereof.

[0194] (C13) In any of the systems shown as (C1) to (C12), the fluid includes human saliva.

[0195] (C14) In any of the systems denoted as (C1) to (C13), the fluid comprises a biological fluid obtained from a human or an animal.

[0196] (C15) In any of the systems described in (C1) to (C14), the signal processing circuit includes a memory to store a kernel matrix having a plurality of cells. Each of the plurality of cells is indexed by a first index value among a plurality of first index values ​​and a second index value among a plurality of second index values. The first index value uniquely identifies one of the first plurality of pixel detectors and the second plurality of pixel detectors. The second index value uniquely identifies one of a plurality of reference wavelengths. Each cell stores the spectral response of the one of the first plurality of pixel detectors and the second plurality of pixel detectors identified by the first index value of each cell, the spectral response being located at the one of the plurality of reference wavelengths identified by the second index value of each cell. The signal processing circuit is also programmed to deconvolve the first raw spectral value array and the second raw spectral value array by performing least squares regression based on at least the first raw spectral value array and the second raw spectral value array and the kernel matrix.

[0197] (C16) In the system shown in (C15), the signal processing circuit is programmed to perform least squares regression with Tikhonov regularization.

[0198] (C17) In any of the systems shown in (C15) and (C16), at least one of the reference wavelengths is within the first wavelength range, and at least one of the reference wavelengths is within the second wavelength range.

[0199] (C18) In any of the systems shown in (C1) to (C17), the signal processing circuit includes a field programmable gate array.

[0200] (C19) In any of the systems shown in (C1) to (C18), the signal processing circuit includes: a processor; a memory that is communicatively coupled to the processor; and a deconvolution engine, which is implemented as machine-readable instructions that are stored in the memory and that, when executed by the processor, control the signal processing circuit to deconvolve the first original spectral value array and the second original spectral value array to obtain the deconvolved spectral value array.

[0201] (C20) In the system shown in (C19), the memory stores a kernel matrix having a plurality of cells. Each cell of the plurality of cells is indexed by a first index value of a plurality of first index values ​​and a second index value of a plurality of second index values. The first index value uniquely identifies one of the first plurality of pixel detectors and the second plurality of pixel detectors. The second index value uniquely identifies one of a plurality of reference wavelengths. Each cell stores the spectral response of the one of the first plurality of pixel detectors and the second plurality of pixel detectors identified by the first index value of each cell, and the spectral response is located at the one of the plurality of reference wavelengths identified by the second index value of each cell. The machine-readable instructions that control the signal processing circuit to perform deconvolution when executed by the processor include machine-readable instructions that control the signal processing circuit to perform least squares regression based on at least the first raw spectral value array and the second raw spectral value array and the kernel matrix when executed by the processor.

[0202] (C21) In the system shown in (C20), the machine-readable instructions that control the signal processing circuit to perform least squares regression when executed by the processor include machine-readable instructions that control the signal processing circuit to perform least squares regression with Tikhonov regularization when executed by the processor.

[0203] (C22) In the systems shown in (C20) and (C21), at least one of the reference wavelengths is within the first wavelength range, and at least one of the reference wavelengths is within the second wavelength range.

[0204] (C23) In any of the systems denoted as (C1) to (C22), the signal processing circuit is further programmed to feed the deconvolved spectral value array into the trained machine learning model by transmitting the deconvolved spectral value array to an external computer system, the external computer system feeding the deconvolved spectral value array into the trained machine learning model to obtain the indication. The signal processing circuit is further programmed to receive the indication from the external computer system.

[0205] (D1) A method includes: transmitting broadband light through a fluid confined within a test box; dispersing a first wavelength range of the broadband light onto a first plurality of pixel detectors forming a first detector array after the broadband light passes through the test box; dispersing a second wavelength range of the broadband light onto a second plurality of pixel detectors forming a second detector array after the broadband light passes through the test box; reading a first raw spectral value array from the first detector array; reading a second raw spectral value array from the second detector array; deconvolving the first raw spectral value array and the second raw spectral value array to generate a deconvolved spectral value array based on the spectral response of each pixel detector in the first plurality of pixel detectors and the second plurality of pixel detectors; feeding the deconvolved spectral value array into a trained machine learning model, the machine learning model processing the deconvolved spectral value array to obtain an indication of the presence of one or more components in the fluid; and outputting the indication.

[0206] (D2) In the method shown in (D1), the method further includes merging the first original spectral value array and the second original spectral value array into an original spectral value composite array. The deconvolution includes deconvolving the original spectral value composite array.

[0207] (D3) In the method shown in (D1), the deconvolution includes deconvolving the first original spectral value array to obtain a first deconvolved spectral value local array and deconvolving the second original spectral value array to obtain a second deconvolved spectral value local array. The method also includes merging the first deconvolved spectral value local array and the second deconvolved spectral value local array to obtain the deconvolved spectral value array.

[0208] (D4) In any of the methods shown in (D1) to (D3), the method also includes collecting the fluid.

[0209] (D5) In the method shown in (D4), the collecting does not include adding any reagent to the fluid.

[0210] (D6) In any of the methods denoted as (D4) and (D5), the collecting includes injecting at least a portion of the fluid into the test cartridge and placing the test cartridge in the cartridge holder.

[0211] (D7) In any of the methods shown in (D4) to (D6), the collecting includes collecting fluid from a human patient.

[0212] (D8) In the method shown in (D7), the collection includes saliva from the human patient.

[0213] (D9) In any of the methods shown in (D6) and (D7), the method also includes diagnosing the human patient with a disease based on the indication.

[0214] (D10) In the method shown in (D9), the method also includes providing therapeutic intervention for treating the disease to the human patient.

[0215] (D11) In the method denoted as (D10), the therapeutic intervention is a surgical procedure, a non-surgical medical procedure, a prescription of one or more medications, or a combination thereof.

[0216] (D12) In any of the methods shown in (D1) to (D11), the deconvolution includes performing least squares regression based on at least the first raw spectral value array and the second raw spectral value array and a kernel matrix comprising multiple cells; each of the multiple cells is indexed by a first index value among a plurality of first index values ​​and a second index value among a plurality of second index values; the first index value uniquely identifies one of the first plurality of pixel detectors and the second plurality of pixel detectors; the second index value uniquely identifies one of a plurality of reference wavelengths; and each cell stores the spectral response of the one of the first plurality of pixel detectors and the second plurality of pixel detectors identified by the first index value of each cell, and the spectral response is located at the one of the multiple reference wavelengths identified by the second index value of each cell.

[0217] (D13) In the method shown in (D12), performing least squares regression includes performing least squares regression with Tikhonov regularization.

[0218] (D14) In any of the methods shown in (D1) to (D13), the feeding includes transmitting the deconvolved spectral value array to an external computer system, wherein the external computer system feeds the deconvolved spectral value array into the trained machine learning model. The feeding also includes receiving the indication from the external computer system.

[0219] (E1) A system includes: a first light source operable to emit a first broadband light beam; a second light source operable to emit a second broadband light beam; and a cartridge holder shaped to receive a test cartridge of a confined fluid. The cartridge holder is positioned so that the first broadband light beam and the second broadband light beam pass through the test cartridge when emitted by the respective first and second light sources. The system also includes a first spectrometer, in response to receiving the first broadband light beam after the first broadband light beam has passed through the test cartridge, the first spectrometer disperses the first broadband light beam onto a first plurality of pixel detectors forming a first detector array. The system also includes a second spectrometer, in response to receiving the second broadband light beam after the second broadband light beam has passed through the test cartridge, the second spectrometer disperses the second broadband light beam onto a second plurality of pixel detectors forming a second detector array. The system also includes a digitizer electrically connected to the first detector array and the second detector array. The digitizer is operable to read a first array of raw spectral values ​​from the first detector array and a second array of raw spectral values ​​from the second detector array. The system also includes a signal processing circuit that is programmed to deconvolute the first raw spectral value array and the second raw spectral value array to obtain a deconvolved spectral value array based on the spectral response of each of the first plurality of pixel detectors and the second plurality of pixel detectors; feed the deconvolved spectral value array into a trained machine learning model, which processes the deconvolved spectral value array to obtain an indication of the presence of one or more components in the fluid; and output the indication.

[0220] (F1) A method includes constructing a kernel matrix of a spectrometer. The kernel matrix includes a plurality of cells. Each cell of the plurality of cells is indexed by a first index value of a plurality of first index values ​​and a second index value of a plurality of second index values. The first index value uniquely identifies one of a plurality of pixel detectors of the spectrometer. The second index value uniquely identifies one of a plurality of reference wavelengths. The construction includes, for each of the plurality of pixel detectors: constructing a spectral response function based on at least (i) one of a plurality of center wavelengths corresponding to the each pixel detector and (ii) one of a plurality of spectral widths corresponding to the each pixel detector; evaluating the spectral response function at the plurality of reference wavelengths to obtain a pixel response vector; and inserting the pixel response vector into the kernel matrix.

[0221] (F2) In the method denoted as (F1), the method further includes downloading the plurality of reference wavelengths from a memory of the spectrometer.

[0222] (F3) In any of the methods denoted as (F1) and (F2), the method further includes setting all of the plurality of spectral widths to the same value.

[0223] (F4) In any of the methods shown in (F1) to (F3), constructing the spectral response function includes constructing a Gaussian function, which is centered on one of the multiple central wavelengths and has a width based on one of the multiple spectral widths.

[0224] (F5) In any of the methods shown in (F1) to (F4), the method also includes uploading the kernel matrix to a memory of the spectrometer.

[0225] (F6) In any of the methods described in (F1) to (F5), the spectrometer includes a spectrophotometer.

[0226] (F7) In any of the methods shown in (F1) to (F6), the multiple reference wavelengths are evenly spaced.

[0227] Changes may be made to the above-described methods and systems without departing from the scope thereof. It should therefore be noted that the matter contained in the above description or shown in the accompanying drawings is to be interpreted as illustrative and not restrictive. The following claims are intended to cover all general and specific features described herein and all statements of the scope of the present methods and systems that, linguistically speaking, can be said to fall therebetween.

Claims

1. A system comprising: a light source operable to emit broadband light; A box support, the box support: A test cartridge shaped to receive a confining fluid; and positioned so that the broadband light passes through the test box when emitted by the light source; a spectrometer, in response to receiving the broadband light after the broadband light passes through the test cell, the spectrometer dispersing the broadband light onto a plurality of pixel detectors of a detector array; a digitizer electrically connected to the detector array and operable to read an array of raw spectral values ​​from the detector array; as well as A signal processing circuit, the signal processing circuit being programmed to: deconvolving the raw array of spectral values ​​based on the spectral response of each of the plurality of pixel detectors to generate an array of deconvolved spectral values; feeding the array of deconvolved spectral values ​​into a trained machine learning model, which processes the array of deconvolved spectral values ​​to obtain an indication of the presence of one or more components in the fluid; and The indication is output.

2. The system of claim 1, wherein: The spectrometer operates within a certain wavelength range; and The broadband light spans the wavelength range.

3. The system of claim 2, wherein the wavelength range includes at least a portion of the visible region of the electromagnetic spectrum.

4. The system of claim 2, wherein the wavelength range comprises 350 nm to 750 nm.

5. The system of claim 2, the wavelength range comprising at least a portion of the near infrared region of the electromagnetic spectrum. The system of claim 2 , wherein the wavelength range includes 900 nm to 1700 nm.

7. The system of claim 1, wherein the spectrometer has a resolution in the range of 0.2 nm to 3 nm (inclusive).

8. The system of claim 1, the spectrometer comprising a spectrophotometer.

9. The system of claim 1, the cartridge holder comprising a cuvette holder, the test cartridge comprising a cuvette.

10. The system of claim 9, wherein the cuvette comprises glass or plastic.

11. The system of claim 1, wherein the fluid comprises human saliva.

12. The system of claim 1, wherein the fluid comprises a biological fluid obtained from a human or an animal.

13. The system of claim 1, wherein: The signal processing circuit comprises a memory storing a kernel matrix comprising a plurality of cells, wherein: Each cell in the plurality of cells is indexed by a first index value in the plurality of first index values ​​and a second index value in the plurality of second index values; The first index value uniquely identifies one of the plurality of pixel detectors; The second index value uniquely identifies one of a plurality of reference wavelengths; and Each cell stores the spectral response of the one of the plurality of pixel detectors identified by the first index value of the each cell, the spectral response being at the one of the plurality of reference wavelengths identified by the second index value of the each cell; and The signal processing circuit is also programmed to deconvolve the raw array of spectral values ​​by performing a least squares regression based on at least the raw array of spectral values ​​and the kernel matrix.

14. The system of claim 13, wherein the signal processing circuit is programmed to perform least squares regression with Tikhonov regularization.

15. The system of claim 1, the signal processing circuit comprising a field programmable gate array.

16. The system of claim 1, wherein the signal processing circuit comprises: processor; a memory communicatively coupled to the processor; as well as A deconvolution engine is implemented as machine-readable instructions that are stored in the memory and that, when executed by the processor, control the signal processing circuit to deconvolve the original array of spectral values ​​to obtain the deconvolved array of spectral values.

17. The system of claim 16, wherein: The memory stores a kernel matrix comprising a plurality of cells, wherein: Each cell in the plurality of cells is indexed by a first index value in the plurality of first index values ​​and a second index value in the plurality of second index values; The first index value uniquely identifies one of the plurality of pixel detectors; The second index value uniquely identifies one of a plurality of reference wavelengths; and Each cell stores the spectral response of the one of the plurality of pixel detectors identified by the first index value of the each cell, the spectral response being at the one of the plurality of reference wavelengths identified by the second index value of the each cell; and The machine readable instructions that, when executed by the processor, control the signal processing circuit to perform deconvolution include machine readable instructions that, when executed by the processor, control the signal processing circuit to perform least squares regression based on at least the array of raw spectral values ​​and the kernel matrix.

18. The system of claim 17, wherein the machine readable instructions that, when executed by the processor, control the signal processing circuit to perform least squares regression include machine readable instructions that, when executed by the processor, control the signal processing circuit to perform least squares regression with Tikhonov regularization.

19. The system of claim 1, wherein the signal processing circuit is further programmed to: feeding the deconvolved spectral value array into the trained machine learning model by transmitting the deconvolved spectral value array to an external computer system, the external computer system feeding the deconvolved spectral value array into the trained machine learning model to obtain the indication; and The indication is received from the external computer system.

20. The system of claim 1, wherein the test cartridge does not contain any reagents.

21. A method comprising: transmitting broadband light through a fluid confined within a test cell; dispersing the broadband light onto a plurality of pixel detectors forming a detector array after the broadband light passes through the test box; reading an array of raw spectral values ​​from the detector array; deconvolving the raw array of spectral values ​​based on the spectral response of each of the plurality of pixel detectors to generate an array of deconvolved spectral values; feeding the array of deconvolved spectral values ​​into a trained machine learning model, which processes the array of deconvolved spectral values ​​to obtain an indication of the presence of one or more components in the fluid; and The indication is output.

22. The method of claim 21, further comprising collecting the fluid.

23. The method of claim 22, wherein said collecting does not include adding any reagents to said fluid.

24. The method of claim 22, wherein said collecting comprises: injecting at least a portion of the fluid into the test cartridge; as well as The test cartridge is placed in the cartridge holder.

25. The method of claim 22, wherein said collecting comprises collecting said fluid from a human patient.

26. The method of claim 25, wherein said collecting comprises collecting saliva from said human patient.

27. The method of claim 25, further comprising diagnosing the human patient with a disease based on the indication.

28. The method of claim 27, further comprising providing said human patient with a therapeutic intervention for treating said disease.

29. The method of claim 28, wherein the therapeutic intervention is a surgical procedure, a non-surgical medical procedure, a prescription of one or more medications, or a combination thereof.

30. The method of claim 21, wherein: The deconvolution comprises performing a least squares regression based on at least the original array of spectral values ​​and a kernel matrix comprising a plurality of elements; Each cell in the plurality of cells is indexed by a first index value in the plurality of first index values ​​and a second index value in the plurality of second index values; The first index value uniquely identifies one of the plurality of pixel detectors; The second index value uniquely identifies one of a plurality of reference wavelengths; and Each cell stores the spectral response of the one of the plurality of pixel detectors identified by the first index value of the each cell, the spectral response being located at the one of the plurality of reference wavelengths identified by the second index value of the each cell.

31. The method of claim 30, wherein said performing least squares regression comprises performing least squares regression with Tikhonov regularization.

32. The method of claim 21, wherein the feeding comprises: transmitting the deconvolved spectral value array to an external computer system, wherein the external computer system feeds the deconvolved spectral value array into the trained machine learning model; as well as The indication is received from the external computer system.

33. A system comprising: a light source operable to emit broadband light; A box support, the box support: a test cartridge shaped to receive a confining fluid; and positioned so that the broadband light passes through the test box when emitted by the light source; a first spectrometer, responsive to receiving a first portion of the broadband light after the broadband light has passed through the test cell, the first spectrometer dispersing a first wavelength range of the first portion of the broadband light onto a first plurality of pixel detectors forming a first detector array; a second spectrometer, responsive to receiving a second portion of the broadband light after the broadband light has passed through the test cell, the second spectrometer dispersing a second wavelength range of the second portion of the broadband light onto a second plurality of pixel detectors forming a second detector array; a digitizer electrically connected to the first detector array and the second detector array, the digitizer operable to: reading a first array of raw spectral values ​​from the first detector array; and reading a second array of raw spectral values ​​from the second detector array; and A signal processing circuit, the signal processing circuit being programmed to: deconvolving the first raw array of spectral values ​​and the second raw array of spectral values ​​based on the spectral response of each of the first plurality of pixel detectors and the second plurality of pixel detectors to generate an array of deconvolved spectral values; feeding the array of deconvolved spectral values ​​into a trained machine learning model, which processes the array of deconvolved spectral values ​​to obtain an indication of the presence of one or more components in the fluid; and The indication is output.

34. The system of claim 33, wherein the signal processing circuit is further programmed to: Merging the first original spectral value array and the second original spectral value array into an original spectral value composite array; and The composite raw spectral value array is deconvolved to obtain the deconvolved spectral value array.

35. The system of claim 33, wherein the signal processing circuit is further programmed to: Deconvolving the first original spectral value array to generate a first deconvolved spectral value local array; Deconvolving the second original spectral value array to generate a second deconvolved spectral value local array; and The first deconvolved spectral value partial array and the second deconvolved spectral value partial array are merged to obtain the deconvolved spectral value array.

36. The system of claim 33, wherein: The first wavelength range includes at least a portion of the visible region of the electromagnetic spectrum; and The second wavelength range includes at least a portion of the near infrared region of the electromagnetic spectrum.

37. The system of claim 33, wherein: The first wavelength range includes 350nm to 700nm; and The second wavelength range includes 900 nm to 1700 nm.

38. The system of claim 33, wherein the first wavelength range and the second wavelength range partially overlap.

39. The system of claim 33, wherein the first wavelength range and the second wavelength range do not overlap.

40. The system of claim 33, wherein: The first spectrometer has a first analytical resolution; and The second spectrometer has a second resolution different from the first resolution.

41. The system of claim 33, each of the first spectrometer and the second spectrometer comprising a spectrophotometer.

42. The system of claim 33, further comprising: Enter the fiber optic cable; a first spectrometer fiber optic cable and a second spectrometer fiber optic cable; a fiber optic collimator that couples the broadband light into the input fiber optic cable after the broadband light passes through the test box; as well as A beam splitter, the beam splitter: splitting the broadband light exiting the input fiber optic cable into a first portion and a second portion; coupling the first portion into the first spectrometer fiber optic cable; and coupling the second portion into the second spectrometer fiber optic cable; wherein (i) the first spectrometer fiber optic cable couples the first portion into the first spectrometer, and (ii) the second spectrometer fiber optic cable couples the second portion into the second spectrometer.

43. The system of claim 33, the cartridge holder comprising a cuvette holder, the cuvette holder positioning a cuvette such that the fluid is illuminated by the broadband light while confined within the cuvette.

44. The system of claim 43, wherein the cuvette comprises glass or plastic.

45. The system of claim 33, wherein the fluid comprises human saliva.

46. ​​The system of claim 33, wherein the fluid comprises a biological fluid obtained from a human or an animal.

47. The system of claim 33, wherein: The signal processing circuit comprises a memory storing a kernel matrix having a plurality of cells, wherein: Each cell in the plurality of cells is indexed by a first index value in the plurality of first index values ​​and a second index value in the plurality of second index values; The first index value uniquely identifies one of the first plurality of pixel detectors and the second plurality of pixel detectors; The second index value uniquely identifies one of a plurality of reference wavelengths; and Each cell stores the spectral response of the one of the first plurality of pixel detectors and the second plurality of pixel detectors identified by the first index value of the each cell, the spectral response being at the one of the plurality of reference wavelengths identified by the second index value of the each cell; and The signal processing circuit is further programmed to deconvolute the first and second raw spectral value arrays by performing a least squares regression based on at least the first and second raw spectral value arrays and the kernel matrix.

48. The system of claim 47, wherein the signal processing circuit is programmed to perform least squares regression with Tikhonov regularization.

49. The system of claim 47, wherein: At least one of the reference wavelengths is within the first wavelength range; and At least one of the reference wavelengths lies within the second wavelength range.

50. The system of claim 33, the signal processing circuit comprising a field programmable gate array.

51. The system of claim 33, wherein the signal processing circuit comprises: processor; a memory communicatively coupled to the processor; as well as A deconvolution engine is implemented as machine-readable instructions that are stored in the memory and that, when executed by the processor, control the signal processing circuit to deconvolve the first raw spectral value array and the second raw spectral value array to obtain the deconvolved spectral value array.

52. The system of claim 51, wherein: The memory stores a kernel matrix comprising a plurality of cells, wherein: Each cell in the plurality of cells is indexed by a first index value in the plurality of first index values ​​and a second index value in the plurality of second index values; The first index value uniquely identifies one of the first plurality of pixel detectors and the second plurality of pixel detectors; The second index value uniquely identifies one of a plurality of reference wavelengths; and Each cell stores the spectral response of the one of the first plurality of pixel detectors and the second plurality of pixel detectors identified by the first index value of the each cell, the spectral response being at the one of the plurality of reference wavelengths identified by the second index value of the each cell; and The machine readable instructions that, when executed by the processor, control the signal processing circuit to perform deconvolution include machine readable instructions that, when executed by the processor, control the signal processing circuit to perform least squares regression based on at least the first and second arrays of raw spectral values ​​and the kernel matrix.

53. The system of claim 52, wherein the machine readable instructions that, when executed by the processor, control the signal processing circuit to perform least squares regression include machine readable instructions that, when executed by the processor, control the signal processing circuit to perform least squares regression with Tikhonov regularization.

54. The system of claim 52, wherein: At least one of the reference wavelengths is within the first wavelength range; and At least one of the reference wavelengths lies within the second wavelength range.

55. The system of claim 33, wherein the signal processing circuit is further programmed to: feeding the deconvolved spectral value array into the trained machine learning model by transmitting the deconvolved spectral value array to an external computer system, the external computer system feeding the deconvolved spectral value array into the trained machine learning model to obtain the indication; and The indication is received from the external computer system.

56. A method comprising: transmitting broadband light through a fluid confined within a test cell; dispersing a first wavelength range of the broadband light onto a first plurality of pixel detectors forming a first detector array after the broadband light passes through the test cell; dispersing a second wavelength range of the broadband light onto a second plurality of pixel detectors forming a second detector array after the broadband light passes through the test cell; reading a first array of raw spectral values ​​from the first detector array; reading a second array of raw spectral values ​​from the second detector array; deconvolving the first raw spectral value array and the second raw spectral value array to generate a deconvolved spectral value array based on the spectral response of each pixel detector in the first plurality of pixel detectors and the second plurality of pixel detectors; feeding the array of deconvolved spectral values ​​into a trained machine learning model, which processes the array of deconvolved spectral values ​​to obtain an indication of the presence of one or more components in the fluid; and The indication is output.

57. The method of claim 56, It also includes merging the first original spectral value array and the second original spectral value array into an original spectral value composite array; The deconvolution comprises deconvolving the original composite array of spectral values.

58. The method of claim 56, wherein: The deconvolution comprises: Deconvolving the first original spectral value array to obtain a first deconvolved spectral value local array; and The second original spectral value array is deconvolved to obtain a second deconvolved spectral value local array; and the method further comprises merging the first deconvolved spectral value local array and the second deconvolved spectral value local array to obtain the deconvolved spectral value array.

59. The method of claim 56, further comprising collecting the fluid.

60. The method of claim 59, wherein said collecting does not include adding any reagents to said fluid.

61. The method of claim 59, wherein said collecting comprises: injecting at least a portion of the fluid into the test cartridge; as well as The test cartridge is placed in the cartridge holder.

62. The method of claim 59, wherein said collecting comprises collecting said fluid from a human patient.

63. The method of claim 62, wherein said collecting comprises collecting saliva from said human patient.

64. The method of claim 62, further comprising diagnosing the human patient with a disease based on the indication.

65. The method of claim 64, further comprising providing said human patient with a therapeutic intervention for treating said disease.

66. The method of claim 65, wherein the therapeutic intervention is a surgical procedure, a non-surgical medical procedure, prescription of one or more medications, or a combination thereof.

67. The method of claim 56, wherein: The deconvolution comprises performing a least squares regression based on at least the first raw spectral value array and the second raw spectral value array and a kernel matrix comprising a plurality of elements; Each cell in the plurality of cells is indexed by a first index value in the plurality of first index values ​​and a second index value in the plurality of second index values; The first index value uniquely identifies one of the first plurality of pixel detectors and the second plurality of pixel detectors; The second index value uniquely identifies one of a plurality of reference wavelengths; and Each cell stores the spectral response of the one of the first plurality of pixel detectors and the second plurality of pixel detectors identified by the first index value of the each cell, the spectral response being located at the one of the plurality of reference wavelengths identified by the second index value of the each cell.

68. The method of claim 67, wherein said performing least squares regression comprises performing least squares regression with Tikhonov regularization.

69. The method of claim 56, wherein the feeding comprises: transmitting the deconvolved spectral value array to an external computer system, wherein the external computer system feeds the deconvolved spectral value array into the trained machine learning model; as well as The indication is received from the external computer system.

70. A system comprising: a first light source operable to emit a first broadband light beam; a second light source operable to emit a second broadband light beam; A box support, the box support: A test cartridge shaped to receive a confining fluid; and positioned so that the first broadband light beam and the second broadband light beam pass through the test box when emitted by the respective first and second light sources; a first spectrometer, in response to receiving the first broadband beam after the first broadband beam has passed through the test cell, the first spectrometer dispersing the first broadband beam onto a first plurality of pixel detectors forming a first detector array; a second spectrometer, in response to receiving the second broadband beam after the second broadband beam has passed through the test cell, the second spectrometer dispersing the second broadband beam onto a second plurality of pixel detectors forming a second detector array; a digitizer electrically connected to the first detector array and the second detector array, the digitizer operable to: reading a first array of raw spectral values ​​from the first detector array; and reading a second array of raw spectral values ​​from the second detector array; as well as A signal processing circuit, the signal processing circuit being programmed to: deconvolving the first raw array of spectral values ​​and the second raw array of spectral values ​​based on a spectral response of each of the first plurality of pixel detectors and the second plurality of pixel detectors to obtain a deconvolved array of spectral values; feeding the array of deconvolved spectral values ​​into a trained machine learning model, which processes the array of deconvolved spectral values ​​to obtain an indication of the presence of one or more components in the fluid; and The indication is output.

71. A method comprising: Construct the kernel matrix of the spectrometer, where: The core matrix includes a plurality of units, wherein: Each cell in the plurality of cells is indexed by a first index value in the plurality of first index values ​​and a second index value in the plurality of second index values; The first index value uniquely identifies one of a plurality of pixel detectors of the spectrometer; and The second index value uniquely identifies one of a plurality of reference wavelengths; wherein the constructing comprises, for each pixel detector in the plurality of pixel detectors: constructing a spectral response function based on at least (i) one of a plurality of center wavelengths corresponding to each of the pixel detectors and (ii) one of a plurality of spectral widths corresponding to each of the pixel detectors; evaluating the spectral response function at the plurality of reference wavelengths to obtain a pixel response vector; and The pixel response vector is inserted into the kernel matrix.

72. The method of claim 71 further comprising downloading the plurality of reference wavelengths from a memory of the spectrometer.

73. The method of claim 71 further comprising setting all of the plurality of spectral widths to the same value.

74. The method of claim 71, wherein said constructing said spectral response function comprises constructing a Gaussian function centered at said one of said plurality of central wavelengths and having a width based on said one of said plurality of spectral widths.

75. The method of claim 71 further comprising uploading the kernel matrix to a memory of the spectrometer.

76. The method of claim 71, wherein the spectrometer comprises a spectrophotometer.

77. The method of claim 71, wherein the plurality of reference wavelengths are evenly spaced.

78. A system or method for detecting disease based on the spectroscopic properties of a biological fluid without the need for reagents.

Images