Predictive diagnostic tests for early detection and monitoring of diseases

Abstract

Methods and systems for the detection and diagnosis of diseases, including cancer. [Solution] A patient's sample is analyzed by absorption spectroscopy in the near-infrared and mid-infrared ranges to create a spectral signature, and the sample is obtained from the patient. This signature is processed by a computational engine using one or more machine learning techniques to determine whether the spectral signature indicates the presence of a disease, including one or more cancers. The embodiment operates outside the conventionally accepted wavelength range, facilitating faster, simpler, and more reliable testing.

Description

Technical Field

[0001] This application generally relates to the field of diagnostic testing, and more particularly to diagnosis by spectroscopic analysis.

Background Art

[0002] Cancer remains a significant global health problem and is the second leading cause of death in the United States. Early cancer detection is still the best weapon against this disease to enable better treatment options and patient outcomes. Current technologies for cancer screening and early detection are still risky, costly, and time-consuming to help win the battle against cancer. Cancer screening is only recommended for high-risk groups based on specific age ranges, smoking history, alcohol consumption, and potential environmental exposures. Cancer incidence is increasing even in populations that do not meet these criteria, such as populations with human papillomavirus-related head and neck cancers. Cancer tests and procedures used for cancer screening and diagnosis include radiation imaging, endoscopy, biopsy, and cytology.

[0003] Tests such as computed tomography (CT), X-ray, and positron emission tomography (PET) involve the risk of radiation exposure that can cause cancer in healthy individuals. Endoscopy and biopsy are invasive and can cause discomfort. These methods of detecting cancer in its early stages depend on individual actions (such as scheduling mammograms, colonoscopies, cervical cytology, etc.), so discomfort and risks can create anxiety and aversion to undergoing tests promptly. In addition to being invasive, histopathological diagnosis is associated with a high probability of false negatives and false positives due to misdiagnosis. This is partly due to some types of tumors being histologically similar and also due to poor cellular differentiation, making it difficult to identify the origin of the tissue.

[0004] While various optical spectroscopy methods have begun to be used for the early diagnosis of cancer, few have shown promising results. Fourier transform spectroscopy (or FTIR) detection systems have been studied, and although many cancer biomarkers often have low sensitivity and specificity in the infrared ("IR") or near-infrared region (e.g., 900–3080 nm), systems have been used with some bright prospects by using high wavelengths across this significant wavelength range. However, FTIR spectroscopy has shown bright prospects due to the fact that it offers a simple, rapid, relatively accurate, inexpensive, non-destructive, and automation-friendly method compared to existing screening, diagnostic, management, and monitoring methods. Infrared spectroscopy is an important tool for screening and detection of cancer and other diseases because it can help improve patient outcomes by informing decision-making and providing early diagnosis. An example of FTIR spectroscopy is described in Kenneth A. Kristoffersen et al., "Fourier-transform infrared spectroscopy for monitoring proteolytic reactions using dry-films treated with trifluoroacetic acid," 10 SCI REP 7844 (2020) (available at https: / / doi.org / 10.1038 / s41598-020-64583-3).

[0005] Near-infrared (NIR) spectroscopy is not a new technique for incorporating cancer screening, but recently published research recommends the use of higher wavelengths and combined techniques for measuring multiple types of samples. The most frequently used range for NIR spectroscopy in cancer detection has been 600–1100 nm. Longer wavelengths tend to be largely absorbed by water, which makes it even more difficult to see any differences. Water is optically transparent but is highly absorbent below 200 nm, and likewise absorbs some NIR as well as more mid-infrared and far-infrared light.

[0006] Additional attempts have been made to utilize liquid biopsy as a means of non-invasive cancer detection. An example of liquid biopsy is shown in U.S. Patent No. 10,288,615, which describes the detection of cancer using blood. Furthermore, studies and disclosures have been presented that measure the water content in normal cells compared to cancer cells. An example of NIR spectroscopy techniques is shown, for example, in U.S. Patent No. 7,706,862, which describes NIR spectral optical imaging using water absorption wavelengths. Near-infrared (NIR) spectral optical imaging at one or more major water absorption wavelengths is used to identify differences in water content between areas of cancer or precancerous tissue and areas of normal tissue. Tissue areas in late-stage cancer have a higher water content than normal tissue.

[0007] The primary water absorption "fingerprint" wavelengths include at least one of 980nm, 1195nm, 1456nm, 1944nm, 2880nm to 3360nm, and 4720nm. Within the range of 400nm to 6000nm, at least one reference wavelength with low or no water absorption, e.g., 4500nm, 2230nm, 1700nm, 1300nm, 1000nm, and 800nm, are used to create reference images to elicit comparisons with images taken at the primary water absorption wavelengths. NIR can be used to measure water absorption by cells, but it is not the most reliable technique and does not produce clear results every time. Because water absorbs a large portion of wavelengths, imaging can be difficult and accuracy is limited. Molecular fingerprinting involves looking for differences in specific biomarkers that constitute a disease fingerprint and in given disease biomarkers. For example, a comprehensive approach to identifying the molecular fingerprint of cancer requires determining not only the presence of cancer but also the type of cancer. Fingerprinting is based on spectral signatures using broad-spectrum spectroscopy, a concept well supported by previous research. By using infrared (IR) spectroscopy, differences in serum components can be recorded to generate unique spectral signatures specific to different health conditions. However, any previous studies using similar techniques have focused on longer wavelength IR absorbances to analyze serum biomolecules. Therefore, there is a need for techniques for non-invasive, low-cost, and low-risk early cancer detection. [Prior art documents] [Patent Documents]

[0008] [Patent Document 1] U.S. Patent No. 10,288,615 [Patent Document 2] U.S. Patent No. 7,706,862 [Non-patent literature]

[0009] [Non-Patent Document 1] Kenneth A. Kristoffersen et al., “Fourier-transform infrared spectroscopy for monitoring proteolytic reactions using dry-films treated with trifluoroacetic acid,” 10 SCI REP 7844 (2020) [Overview of the Initiative] [Means for solving the problem]

[0010] This disclosure provides a method and system for diagnosing diseases, which involves performing absorption spectroscopy in the near-infrared to mid-infrared (NIR-MIR) range on a patient sample and analyzing the resulting spectral signature to determine the presence of diseases such as cancer.

[0011] According to one aspect of the present disclosure, a method for diagnosing a disease includes the step of analyzing a sample obtained from a patient by absorption spectroscopy in the NIR-MIR range to create a spectral signature. The spectral signature is received by a processor. The processor determines whether the spectral signature consequently indicates the presence of a disease and outputs the result. The method may optionally include providing the patient with a treatment to treat the disease when the result indicates the presence of a disease.

[0012] According to another aspect of the present disclosure, a system for diagnosing the presence of a disease includes a memory and a processor configured to execute instructions stored in the memory to receive a spectral signature made by absorption spectroscopy in the NIR and / or MIR range of a sample from a patient, determine whether the spectral signature consequently indicates the presence of a disease, and output the result.

[0013] In the embodiment, the sample may be a lysate sample. A lysate sample can be obtained by adding a solubilizing solution or a homogenizing solution to a serum sample. The solubilizing solution or homogenizing solution may be added to the serum sample in a 1:1 ratio. In the embodiment, at least one of a photobinding solution or a proteolytic reagent is added to the lysate sample. In the embodiment, any combination of the solubilizing solution, homogenizing solution, photobinding solution, and proteolytic reagent may be added to the lysate sample.

[0014] The embodiment may further include drying the sample on an IR reflective sampling card. The card may be coated with aluminum or other non-IR absorbing material.

[0015] In one embodiment, the processor can determine whether a spectral signature indicates the presence of a disease by providing the spectral signature to a computation engine that includes a model architecture and one or more model parameters. The computation engine can execute computation algorithms, such as machine learning algorithms, configured to provide results based on the spectral signature, the model architecture, and one or more model parameters.

[0016] In the embodiment, the computation engine can receive feedback indicating the accuracy of the results. At least one model parameter may be updated based on the feedback.

[0017] Another aspect of the present disclosure discloses a method for detecting pathogens in a specimen. The method includes the steps of receiving a whole blood specimen from a patient into a coagulation cuvette, operating the coagulation cuvette to release a serum specimen into an analysis chamber, inserting the cuvette into a spectrophotometer using near-infrared and / or mid-infrared spectra, determining whether a pathogen is present in the serum specimen, and outputting a result indicating whether a pathogen has been detected.

[0018] According to another aspect of the present disclosure, a system for detecting pathogens in a specimen includes a coagulation cuvette, a spectrophotometer using near-infrared and / or mid-infrared spectra, and a processor associated with the spectrophotometer and configured to perform analysis of spectrophotometric signatures output by the spectrophotometer.

[0019] According to another aspect of the present disclosure, a serum isolation cuvette includes a serum analysis chamber comprising a sample container and one or more serum channels above the sample container, and a coagulation chamber comprising a coagulant for coagulating a sample introduced into the coagulation chamber, a coagulation strainer for removing all coagulated cells from the sample, and a channel plug. The channel plug and serum channels can form a releaseable seal between the coagulation chamber and the analysis chamber so that sample serum can flow into the analysis chamber.

[0020] Embodiments of this disclosure provide artificial intelligence and machine learning-driven optical molecular sensing systems and methods for detecting diseases, including early cancer detection. Embodiments include a portable, power-integrated optical sensing device, a disposable sampling tube, or a microfluidic cassette (microcubet) along with an optical sensing cocktail solution. Data is collected and analyzed by an application having a predictive training database (locally and cloud-based) on a smart device or computer. The system can be applied for the analysis of cancer-specific biomolecules performed directly in serum. Biomolecular analysis requires relatively small sample volumes (approximately 50–200 μL), and the analysis is performed without damaging the sample, within a timeframe of 20 minutes or less.

[0021] The above summary is not intended to describe each of the illustrated embodiments or all implementations of the subject matter herein. The following drawings and detailed description illustrate various embodiments in more detail.

[0022] The subject matter of this specification can be more fully understood by considering the following detailed description of various embodiments in conjunction with the accompanying drawings.

Brief Description of the Drawings

[0023] [Figure 1] A flowchart showing a method for detecting a disease using a centrifuge and serum according to one embodiment. [Figure 2] A flowchart showing a method for detecting a disease using a serum cocktail according to one embodiment. [Figure 3] A flowchart showing a method for detecting a disease using a dedicated cuvette according to one embodiment. [Figure 4] A flowchart showing a method for detecting a disease using a lysate according to one embodiment. [Figure 5] A schematic diagram showing a computing engine according to one embodiment. [Figure 6] A schematic diagram showing a method for training a computing engine according to one embodiment. [Figure 7] A diagram of a normalized spectrum showing the difference in spectroscopic output. [Figure 8] A diagram of a normalized spectrum showing the difference in spectroscopic output. [Figure 9] A diagram of a normalized spectrum showing the difference in spectroscopic output. [Figure 10] A diagram of a normalized spectrum showing the difference in spectroscopic output. [Figure 11A] A chart showing an exemplary output according to one embodiment. [Figure 11B] A chart showing an exemplary output according to one embodiment.

Modes for Carrying Out the Invention

[0024] While various embodiments may accommodate various modifications and alternative forms, their details are shown as examples in the drawings and described in detail. However, it should be understood that the claimed invention is not limited to the specific embodiments described. In contrast, the invention encompasses all modifications, equivalents, and alternatives that fall within the spirit and scope of the subject matter defined by the claims.

[0025] Non-invasive diagnosis and early detection of diseases such as cancer remain a major challenge in healthcare today. New and emerging non-invasive tests, including biomarker / genetic testing, liquid biopsies, and breath biopsies, are being developed. However, high costs, lengthy testing procedures, and low positive test accuracy make these tests unsuitable for routine testing. The systems and methods of this disclosure are designed to revolutionize liquid biopsy-based diagnostics, being performed directly, rapidly, and at low cost in serum or plasma, eliminating the need for multiple lengthy sample processing steps.

[0026] Liquid biopsy-based diagnostics can offer high sensitivity and accuracy for disease detection. Current liquid biopsy techniques for cancer diagnosis include the analysis of circulating tumor cells (CTCs) and cell-free circulating tumor (CT) DNA, mRNA, and microRNA (miRNA) in blood samples. Others include the isolation and analysis of extracellular vesicle / exosome biomarkers from plasma or serum. These techniques rely on the analysis of a predetermined set of biomarkers based on flow cytometry, next-generation sequencing polymerase chain reaction (PCR) and related chemistry-based detection, as well as the use of protein / glycan detection. However, these techniques are still in their early stages, and their success is limited by several factors.

[0027] Firstly, due to the heterogeneity of cancer, a single liquid biopsy platform cannot consistently and reliably measure a predetermined set of biomarkers for early cancer detection. Secondly, sample analysis, which requires multiple steps of preparation, manipulation, long-term chemical reactions, detection, and data analysis, is prone to technical and human errors, not to mention the high cost to technological development and utility in clinical settings. Finally, current liquid biopsy techniques require large sample volumes for analysis, and samples cannot be reused for multiple tests. The solution to these problems is an innovative, non-destructive, and broad molecular fingerprint-based liquid biopsy technique that goes straight from serum to detection, producing results within 10-30 minutes without lengthy sample processing. Furthermore, samples can be reused for multiple tests.

[0028] The optical sensing described herein is based on spectrophotometric analysis of cancer-specific molecular "fingerprints" generated by a complex mixture of biomolecules present in serum. This technique operates on the principle that blood samples from patients with advanced cancer contain biomolecules with different structural compositions than samples from healthy individuals. These differences include biochemical components, proteins, circulating cell-free DNA / RNA / miRNA, and glycans (carbohydrates) contained within lipid-coated extracellular vesicles (EVs), such as exosomes that are not only soluble in serum but also present in serum.

[0029] Exosomes are small, membrane-bound extracellular molecules (EVs) secreted by immune cells for intercellular communication, such as those in tissues and immune cells responding to the presence of cancer. While smaller than other EVs found in serum, such as apoptotic bodies (released from dying cells) and microvesicles, they contain biological components crucial to cancer cells. More importantly, cancer cells generally secrete more exosomes and process more biomolecules than healthy cells. Overall cancer-related exosome content can be distinguished from healthy exosomes because they are used to mediate communication between other cancer cells within the tumor microenvironment (or ecological niche) to modulate and suppress immune responses and to establish ecological niches that support cancer growth and spread. In addition, serum biomolecules, particularly the densely packed cargo within exosomes, form distinct intermolecular interactions and chemical bonds, resulting in molecular fingerprints specific to different physiological / pathological conditions.

[0030] Unlike other techniques that focus specifically on DNA, mRNA, miRNA, or exosomes, this disclosure detects a complex mixture of all components present in serum and larger extracellular matrix (EVs) and exosomes. This is beneficial because, at the very least, cancer-related components are also found in larger microvesicles as well as in apoptotic bodies of dying cancer cells. By measuring the biochemical and physical properties of serum and EV components, differences in molecular fingerprints can be associated with the presence of different types of cancer. The fingerprints are based on spectral signatures using broad-spectrum spectroscopy. This concept is well supported by multiple studies reported in the literature and our own preliminary data. By using infrared (IR) spectroscopy, differences in serum molecular composition can be recorded to generate unique spectral signatures or fingerprints specific to different health conditions.

[0031] This disclosure provides a fast, low-cost, and low-risk screening and monitoring method for early signs of diseases such as cancer, using a small amount of blood sample. Because this disclosure utilizes optical sensing technology, the sample is not destroyed and can be used for other tests without requiring additional blood collection. The cancer sensing strategy disclosed herein explores differences in a broad, biomarker-independent cancer fingerprint and combines them with signals specific to a given cancer biomarker to generate over 1000 data points in a single readout (approximately 2 seconds per scan). This technique enables the early detection of different types of cancer from a single test using a single sample.

[0032] The technology of this disclosure measures serum for molecular fingerprinting by using absorption spectra from visible and near-infrared (VIS / NIR) to mid-infrared (MIR) light as a base component, combined with a modulated NIR / MIR spectrum and a fluorescence spectrum specific to a given biomarker. This allows the sensing device embodying this disclosure to be compact, portable, and battery-powered. In addition to the base spectroscopic signature and biomarker-specific signals that define the molecular fingerprint, we also measure modulated absorption spectra using specific compounds, peptides, and antibodies designed to bind to specific sets of cancer-specific proteins, nucleic acids, and carbohydrates, to create a broad molecular fingerprint for a multi-cancer detection platform and improve the accuracy of early cancer detection.

[0033] Throughout this specification, wavelength or wavenumber ranges are referred to as “mid-IR” or “near-IR.” Depending on the context, these terms may have different meanings in various technical disciplines. However, for the purposes of this document, please understand that near-IR or NIR refers to light with wavelengths between approximately 600 nm and approximately 2500 nm, and in most applications described herein, between 600 nm and 1100 nm. Mid-IR or MIR refers to light with wavelengths between approximately 1250 nm and approximately 25,000 nm, and in most applications described herein, between 2500 nm and 25,000 nm. “Infrared light source” refers to one or more light sources that produce or emit radiation within the infrared wavelength range, and for example, “infrared light source” may include wavelengths within mid-IR (2 to 2.5 microns). Infrared light sources may generate radiation over most of a small area of ​​these wavelengths, or they may have a tuning range which is a subset of one of the wavelength ranges, or they may provide radiation over multiple discrete wavelength ranges, for example, 2.5–4 microns or 5–13 microns. The radiation source may be one of many sources, including heat sources or global sources, supercontinuum laser sources, frequency combs, difference frequency generators, sum frequency generators, harmonic generators, optical parametric oscillators (OPOs), optical parametric generators (OPGs), quantum cascade lasers (QCLs), nanosecond, picosecond, femtosecond and attosecond laser systems, CO2 lasers, heated cantilever probes or other minute heaters, and / or any other sources that create a beam of radiation. The source may be narrowband, for example, with a spectral width less than 10 cm⁻¹ or less than 1 cm⁻¹, or broadband, for example, with a spectral width greater than 10 cm⁻¹, greater than 100 cm⁻¹ or greater than 500 cm⁻¹.

[0034] An "infrared absorption spectrum" refers to a spectrum proportional to the wavelength dependence of the infrared absorption coefficient, absorbance, or similar representation of the IR absorption characteristics of a specimen. An example of an infrared absorption spectrum is the absorption measurement produced by a Fourier transform infrared spectrometer (FTIR), i.e., the FTIR absorption spectrum. In general, infrared light is either absorbed (i.e., part of the infrared absorption spectrum), transmitted (i.e., part of the infrared transmission spectrum), or reflected. The reflection or transmission spectrum of the collected light may have different intensities at each wavelength compared to the intensity at that wavelength in the probe light source.

[0035] Terms such as "approximately" or "approximately" are synonyms and are used to indicate that a value modified by these terms has an associated understood range, which may be ±20%, ±15%, ±10%, ±5%, or ±1%.

[0036] Embodiments of this disclosure relate to disease detection devices and procedures, as shown overall in flow path 100 in Figure 1. While this disclosure focuses, as an example, on cancer as a diagnosis of interest, the principles, devices, and methods of this disclosure may be applicable to the diagnosis of other diseases, such as heart disease, diabetes, and COVID-19.

[0037] A whole blood sample 102 is collected from the patient, centrifuged 104, and serum 106 is extracted. The serum is then optically analyzed 108 to create a spectral signature 110 or "molecular fingerprint" of EVs and exosomes present in the patient's blood. This spectral signature contains unique peaks for each type of molecule present in the serum, including lipids, proteins, carbohydrates, and nucleic acids.

[0038] Although whole blood sample 102 is described as the target sample in this and other examples herein, other samples may be used in embodiments, but are not limited to cerebrospinal fluid or other fluid biopsies, saliva, urine, etc. Analysis of solid or liquefied tissue, or vapor, such as breath, is also conceivable. Whole blood is generally preferred due to the ease of collection and non-invasive nature of whole blood samples, as well as the holistic representation of blood to cells in various systems of the body.

[0039] By centrifugating the sample 104, whole blood cells are removed from the serum 106, preventing larger whole cells from obscuring EVs and exosomes. In some embodiments, centrifugation of the sample may be omitted, thereby allowing the whole blood sample to be analyzed by the spectrophotometer 108.

[0040] In the embodiment, a sensor cocktail solution may be added to serum 106 before analysis by the photosensor 108. The sensor cocktail solution is a mixture of antibodies, peptides, and / or molecular binding reagents against cancer-type specific proteins and biomolecules, including EGF, INHBA, CD44, and other extracellularly secreted / released molecules present in serum associated with different types of cancer. The antibodies and peptides are either unbound for modulated absorption spectral measurements or labeled with fluorescence resonance energy transfer (FRET) fluorescent dye molecules for fluorescence spectral analysis. The addition of the sensor cocktail modifies the output signature according to the presence or absence of the target molecule, providing presence / absence detection of whether cancer-specific molecules are present in the sample. The target antibody may be labeled by any means available in the art, but such modification is unnecessary for target identification under the principles and methods of this disclosure.

[0041] Various possible combinations of cocktail solutions are envisioned, comprised of a wide range of antibodies, peptides, and molecular conjugate reagents known in the art. For example, a general cancer detection solution may identify broad arrays of extracellular proteins (EVs) and exosomes associated with various cancer strains to provide a screen measure to determine whether more targeted testing is appropriate. Other cocktail solutions may be specifically formulated for cell products associated with particular cancer strains, such as lung cancer or breast cancer, or to produce specific outputs characteristic of certain cancer strains.

[0042] The sample may be analyzed by a spectrophotometer 108 to produce a spectral signature 110. While any broad spectrum spectral frequency may be used, in some preferred embodiments, infrared (IR), near-infrared (NIR), or visible (VIS) light spectra may be used. Shorter wavelengths in these ranges may be preferable to allow the use of smaller, more compact spectrophotometers, which can further improve the mobility and availability of the test by additionally enhancing the portability of the required instrument. Shorter wavelength spectrophotometers may also be preferable due to lower power requirements and easier sample preparation, as aqueous samples can be analyzed.

[0043] Despite these advantages provided by the NIR or MIR range, the resulting spectral signature 110 is often complex due to the broad harmonics and the coupling band created by the NIR or MIR absorbance. As will be described in more detail below with reference to Figure 4, the computing engine 600 may be configured to receive the spectral signature 110 and to automatically receive one or more interpretations of the spectral signature 110.

[0044] Figure 2 is a schematic diagram showing possible modes of sample collection and preparation according to the embodiment.

[0045] The first system 200 may include a rapid blood coagulation microcentrifuge tube for blood collection and a high-speed microcentrifuge for serum separation, an optically transparent microcubet for serum and sensor cocktail mixes, a Bluetooth® / Wi-Fi connected optical sensor, and a computer or smart device having a software application that includes or can be communicably connected to a computing engine 600. Thus, system 200 can provide a laboratory-based analytical system.

[0046] The second system 300 may include a serum separation microcuvette designed for collecting blood drops by fingertip prick, which can be done at home for personal use or in other non-laboratory environments. The collected serum can be mixed with a pre-configured sensor cocktail and analyzed in a portable optical sensor, or sent to a laboratory for more comprehensive analysis. Thus, system 300 can provide a more portable analytical system, such as a home system or a mobile system.

[0047] Figure 3 is a schematic diagram showing a serum separation microcubet provided by an embodiment. A patient or other user may have a small, low-power spectrophotometer, such as an NIR or VIS spectrophotometer, which may be privately owned or available in a public place such as a pharmacy or clinic. Such a personal or public machine may be designed to be received directly and analyze the serum separation microcubet. The user may screen themselves for cancer or other diseases of interest by performing a fingertip puncture 302 and placing a few drops of whole blood into the serum separation microcubet 304. The serum separation microcubet is placed in an optical sensor 306 and may be analyzed by an associated processor 308. Output results are generated 310 and may be presented to the user, or the system may be configured to allow the user to send the results directly to a healthcare provider, such as by appropriately contacting the healthcare provider or by integrating with an electronic medical record system.

[0048] The serum separation microcubette 400 includes a coagulation chamber 410 and a serum analysis chamber 420. In embodiments, the microcubette 400 may be pre-configured to support its ease of use function. The coagulation chamber 410 is placed above the serum analysis chamber 420. A user may, for example, place a few drops of blood into the upper coagulation chamber 410 by finger prick. The blood cells to coagulate are captured in the coagulation chamber 410, and the serum can flow down into the lower serum analysis chamber 420 or out by gravity.

[0049] The coagulation chamber 410 includes a self-locking cap 412, a coagulation strainer 414, and a serum channel plug 416. The coagulation chamber 410 is coated with a coagulant so that blood or other biological specimens begin to coagulate when added to the chamber. The self-locking cap 412 may be constructed, for example, as an interference fit to permanently seal the open top of the coagulation chamber 410. The bottom of the coagulation chamber may be releasably sealed by a channel plug 416, etc. The coagulation strainer 414 may be integrated with the cap 412 or other components placed on top of the channel plug 416 to capture coagulating cells so that serum can be released into the analysis chamber 420 below. The channel plug 416 may hold the serum and cells of a specimen in the coagulation chamber 410 until sufficient cells have been removed by the strainer 414 to produce a satisfactory serum specimen. The plug 416 may be released by lifting or twisting the coagulation chamber 410, etc., to allow serum to flow into the analysis chamber 420 below.

[0050] The analysis chamber 420 includes a container 422 and a drain channel 424. The drain channel 424 at the top of the analysis chamber 420 can direct serum toward the container 422 below, resulting in some additional sample filtration. The channel 424 may also be involved in a releaseable seal between the coagulation chamber 410 and the analysis chamber 420. For example, the channel 424 may be connected to a plug 416 of the coagulation chamber to prevent the sample from entering the analysis chamber 420 until the plug 416 is released, for example, by lifting the coagulation chamber 410 and removing the plug 416 from between the channels 424. Once the serum is collected in the container 422, it can be analyzed by spectroscopy or the like. In embodiments, the analysis chamber may also play a role in enabling cocktail mixing by adding a target antibody or peptide on or directly into the container 422 so that the serum mixes with the target substance as the serum collects in the container 422. At least the container 422, and in the embodiment all or any components of the cuvette 400, are formed of an optically transparent material.

[0051] In embodiments, the method of the present disclosure may be used with a lysate, which is shown overall in the flow path 500 in Figure 4. A serum sample 502 is prepared by collecting a whole blood sample from a patient and centrifuging it to remove red blood cells. An EV solubilizing and homogenizing solution is then added to the serum 502 in a 1:1 or optimal ratio to dissolve the EV in the serum 502 to create a lysate 504. A base sample 506 consisting only of the lysate may be used for optical fingerprinting, or the lysate 504 may be modified before optical fingerprinting. A first modified sample 508 may include a photobinding solution added to the lysate for optical fingerprinting. The photobinding solution may include an antibody, a peptide, and other components for cancer type-specific proteins and biomolecules. A second modified sample 510 may include a proteolytic reagent added to the lysate 504. The proteolytic reagent may break down the peptides in the lysate into smaller peptides for optical fingerprinting. In some embodiments, two or more combinations of lysates alone, lysates with a photobinding solution, and lysates with a proteolytic reagent may be used as optical fingerprinting samples.

[0052] Next, the optical fingerprinting sample may be coated onto an attenuated total internal reflection ("ATR") crystal 512 or an infrared ("IR") reflection sampling card 514 and dried. Once the optical fingerprinting sample is dried, Fourier transform infrared spectroscopy (FTIR) is performed using systems and methods known in the art 516. When FTIR is performed on an ATR crystal, an IR beam travels into the ATR crystal having a defined refractive index and passes through the sample on the dried film. The IR beam is refracted and returns into the crystal, and a detector detects the resulting wavelength to create a spectral signature.

[0053] In some embodiments, the optical fingerprinting specimen is dried on an IR reflective sampling card 514, such as an aluminum-coated card. Aluminum is highly reflective and reflects the IR beam, and other coatings, such as those containing one or more materials including gold, indium, tin oxide, and zinc oxide, are considered. FTIR 516 is then performed on the coated IR reflective sampling card, in which the IR beam passes through the card and the dried specimen, reflecting the IR beam detected by the detector, and the reflected beam creates a spectral signature.

[0054] Serum analysis using the systems and methods disclosed herein can produce results within 30 minutes, 20 minutes, or 10 minutes, depending on the system configuration. The disclosure may be advantageously used in resource-constrained environments by utilizing an integrated system consisting of an optical sensor device, either a microcentrifuge or a serum separation microcubet, and a smartphone or laptop computer. After separation from a thrombus, the serum sample placed inside the microcubet is analyzed using the optical sensor device to perform base and modulated absorption spectroscopy. Using the methods and systems disclosed herein, small samples can yield definitive results; for example, samples of 50 μl, 60 μl, 70 μl, 80 μl, or less can be effectively used.

[0055] Figure 5 is a schematic diagram showing the components of a computing engine 600 according to one embodiment. In this embodiment, the computing engine 600 can use one or more computing techniques, including artificial intelligence techniques such as machine learning, to enable efficient interpretation of spectral signatures.

[0056] Many different machine learning algorithms exist, and generally, machine learning algorithms search to approximate an ideal target function f that best maps the input variable x (domain) to the output variable y (range), and therefore, y = f(x).

[0057] Therefore, machine learning algorithms as approximations of f are suitable for providing predictions of y. A supervised machine learning algorithm generates a model to approximate f based on a training dataset, where each of the training datasets is associated with an output y. The supervised algorithm generates a model in which predictions approximate f by a training process that can be formulated based on the output y associated with the training datasets. The training process can be iterated until the model achieves a desired level of accuracy on the training data.

[0058] Other machine learning algorithms do not require training. Unsupervised machine learning algorithms generate models that approximate f by inferring structures, relationships, themes, and / or similarities present in the input data. For example, rules may be extracted from the data, mathematical processes may be applied to systematically reduce redundancy, or the data may be structured based on similarities. Semi-supervised algorithms, such as hybrids of supervised and unsupervised methods, may also be employed.

[0059] In particular, the range y of f could be, among other things, a set of classification schemes that are formally enumerated, extensible, or uncertain, so that domain x is classified for, for example, labeling, categorization; a set of clusters of data where clusters can be determined based on features of domain x and / or intermediate range y'; or a continuous variable such as a single value, a series of values.

[0060] Regression algorithms for machine learning can model f over a continuous range y. Examples of such algorithms typically include least squares regression (OLSR), linear regression, logistic regression, stepwise regression, multivariate adaptive regression spline (MARS), and local estimate scatter plot smoothing (LOESS).

[0061] Clustering algorithms can be used, for example, to infer f and describe hidden structures from data that includes unlabeled data. Such algorithms include, among others, K-means clustering, mixture models, neural networks, and hierarchical clustering. Anomaly detection algorithms may also be employed.

[0062] Classification algorithms address the challenging task of identifying which of a set of categories (range y) a given observation (domain x) belongs to. Such algorithms are typically supervised or semi-supervised, based on a training set of data. Algorithms can include, among others, linear classifiers such as Fisher's linear discriminant, logistic regression, and Naive Bayes classifiers; support vector machines (SVMs) such as least-squares support vector machines; quadratic classifiers; kernel estimation; decision trees; neural networks; and learning vector quantization.

[0063] The computing engine 600 may be a separate software module in the sense that it is single, individual and / or unique, and may be portable in the sense that the computing engine 600 may be stored or transmitted for execution in potentially multiple execution environments, such as physical or virtual computer systems or software platforms that run in computer systems such as runtime environments, operating systems, and platform software.

[0064] The computing engine 600 may include executable computing algorithms 602, such as any of the machine learning algorithms previously described herein or any other suitable machine learning algorithms that are obvious to those skilled in the art. A suitable machine learning algorithm can be configured to run within the scope of the computing engine 600 based on input parameters, such as domain data and / or configuration parameters, as input to the algorithm to produce machine learning results, such as range data and / or other output data. For example, the computing algorithm 602 may be provided as a procedure or function in a software library, as a software object or subroutine. Thus, a machine learning algorithm can be executable to perform a machine learning function which includes either or all of a training phase of behavior for the algorithm to train the computing engine 600 in a supervised or semi-supervised manner, and / or a processing phase of behavior for the algorithm to provide one or more machine learning results.

[0065] The computing engine 600 may further include a data storage device 604 for storing data by algorithms. The data storage device may be a volatile or non-volatile storage device, such as memory. The data storage device 604 may be used to store data required by algorithms, such as machine learning parameters 606, and machine learning data structures, including, in particular, tree data structures, representations of regression analysis data structures, representations of neural network data structures, variables, and any other data that can be stored by machine learning algorithms obvious to those skilled in the art. Thus, the computing engine 600 provides one or more machine learning algorithms and separate encapsulations for or required by such algorithms.

[0066] The computing engine 600 may further include an interface 610 for communicating with the outside world to receive inputs and deliver outputs. For example, machine learning parameters, including configuration information, training data, and machine learning input (domain) information, may be communicated via the input as at least part of the input data 612. Results 614, such as predictions generated by the model, may be communicated via the interface as outputs. Feedback 616 may include additional inputs that can be used during individual training operations of the computing engine 600 or provided as inputs during normal operation of the computing engine 600 to provide iterative learning.

[0067] Interface 610 can receive user input and provide the user with output regarding the configuration of the compute engine 600. Interface 610 may include a mobile application, a web-based application, or any other executable application framework. The interface may reside on, present on, or be accessed by any computing device capable of communicating with various components of the compute engine 600, receiving user input, and presenting output to the user. In embodiments, the interface may reside on or be presented on a smartphone, tablet computer, laptop computer, or desktop computer.

[0068] The computation engine 600 can implement any suitable machine learning algorithm or architecture having any number of layers, such as convolutional layers, activation layers, and pooling layers. Furthermore, the computation engine 600 can be trained by any suitable training method. In one embodiment, the computation engine 600 may include a single-layer neural network, which may be trained using supervised training techniques.

[0069] Training may be performed by targeting known molecules, as well as by providing spectra for known diagnostics and controls. In embodiments, results may be normalized to total absorbance before analysis to enhance the overall clarity and consistency of the results.

[0070] In one embodiment, training data may include spectral signatures 110, which may be labeled with known diagnoses such as healthy (cancer-free), breast cancer, lung cancer, other cancers, or combinations thereof. Thus, the computation engine 600 can classify a given spectral signature input to indicate a diagnosis (along with its likelihood).

[0071] The training data may also include spectral signatures 110 that are labeled based on the presence (or deficiency) of known molecules of interest (lipids, proteins, etc.). Therefore, the computational engine can classify a given spectral signature input to indicate the presence (along with their likelihood) of one or more molecules of interest.

[0072] In addition to the spectral signature 110, in the embodiment, the input to the calculation 112 (whether from the training data or the data being evaluated) may include additional information about the patient, such as demographic data (age, ethnicity, etc.) and health characteristics (height, weight, medical history, medication, etc.). Therefore, the calculation engine 600 can use the additional patient data to determine the output.

[0073] Figure 6 is a flowchart of a method 1000 for training a computation engine 600. In 1002, a labeled dataset spectral signature 110 may be received or provided, and as described above, the dataset may further include additional patient data. In 1004, the labeled dataset may be split into training data and examination data, for example, the training dataset may include a certain percentage, such as 60% of the entries in the labeled dataset. In 1006, a training dataset including relevant known diagnoses, molecular presence, or other factors may be provided to the model. In 1008, the model may generate a first set of parameters by any suitable method.

[0074] In step 1010, part or all of the test dataset may be provided to the model without any known entries. In step 1012, the predictions generated by the model may be compared to the known entries for each input in the test dataset. In step 1012, if the recommendation shows an acceptable amount of error, additional testing may occur by iterating in step 1016. In step 1012, if the recommendation shows an unacceptable amount of error, the classification parameters may be modified before proceeding to the iteration in step 1016. This testing phase may be repeated until the test dataset is exhausted, or until the amount of error steadily reaches a minimum threshold over a set time period, or until some other criterion is met.

[0075] Figures 7 to 10 are charts illustrating examples of recordings made by embodiments of the present disclosure, which are described in detail below. [Examples]

[0076] Evidence of a concept Serum samples of 200 μl per patient were collected by centrifugation of coagulated whole blood. Spectroscopy was performed to measure base and modulated absorption spectra in the range of 403 nm to 724 nm with a resolution (interval) of approximately 3–4 nm. Total absorption spectral measurements were performed continuously for approximately 5–10 seconds, or until the absorption signal stabilized (usually within 20 seconds). The final absorption spectrum for each sample was averaged over the total measurement time. The averaged base absorption spectra of serum samples from four healthy individuals (non-cancer), three lung cancer patients, and two breast cancer patients were individually normalized against the total absorbance value for each sample and plotted for visualization (Figure 7). Distinctive spectra were observed between healthy individuals and lung cancer patients, while the difference between healthy individuals and breast cancer patients was less pronounced but still discernible. [Examples]

[0077] specificity To determine whether serum spectroscopy from cancer patients was a measurement of cancer-related molecular fingerprints, serum samples from lung and breast cancer were mixed with six different healthy serum conditions. If the observed absorption spectrum depended on the total serum contents regardless of cancer, cancer-specific spectroscopy would be masked by non-cancerous healthy serum components, resulting in a spectrum indistinguishable between healthy and cancerous samples. To increase serum complexity, two healthy serum samples were mixed in a 1:1 ratio to produce six different serum combinations. Each cancer serum sample was mixed in a 1:1 ratio with each healthy serum mixture, creating 18 lung cancer samples and 12 breast cancer samples at 50% dilutions for each sample. Cancer samples diluted at 50% within different combinations of healthy serum produced more distinctive absorption spectra compared to the six different healthy serum combinations (Figure 8). These results indicate that molecular fingerprint spectroscopy was specific to the lung and breast cancer samples examined. [Examples]

[0078] modulation The absorption spectrum was investigated to see if it could be modulated using an anti-human epidermal growth factor (EGF) antibody. EGF is a key growth factor produced by cancer cells for tumorigenesis and is present in higher levels in the serum of cancer patients. The hypothesis was that the anti-EGF antibody would bind to the EGF protein present in the serum of cancer patients, altering its overall molecular binding and thus altering the protein's light absorption spectrum. To test this hypothesis, healthy serum and lung cancer serum were mixed with 1 μg of antibody (total 1 vol%) and cultured at ambient temperature for 0, 10, and 20 minutes.

[0079] At culture times of 10 minutes (data not shown) and 20 minutes, a slight increase in the absorption spectrum of healthy serum was observed between 403 nm and 425 nm, but the absorbance of lung cancer serum was specifically modulated at different wavelengths (Chart A in Figures 9 and 10). The absorbance of lung cancer serum shifted upward in the shorter wavelength region at 10 minutes of culture and downward after 20 minutes of culture (Chart B in Figure 10). In contrast to culture times of 0 and 10 minutes, the serum shifted upward in the longer wavelength region after 20 minutes in the presence of EGF antibodies (Chart C in Figure 10). These results suggest that the increased absorbance of light at shorter wavelengths was attributable to the presence of unbound antibodies. When bound to the EGF protein in the serum of cancer patients, the antibodies increased the absorbance of light at longer wavelengths, creating a modulated absorption spectrum. The increase in absorbance appeared to correlate with the increase in wavelength. As part of our technological development, spectroscopy is performed from 724 nm to 950 nm or longer wavelengths to determine whether a further increase in absorbance is observed. [Examples]

[0080] identification To demonstrate whether the system of this disclosure can distinguish lung cancer from breast cancer serum samples and healthy serum samples, a neural network machine learning algorithm was used to train and build a predictive model using normalized data from serum spectroscopy. All samples tested, including undiluted and diluted (up to 64-fold) samples, were included in a one-out cross-validation where one sample was withheld from testing and the remaining samples were used to train the model. The cross-validation process was repeated until all samples were tested. The results of the machine learning cross-validation are visualized in Figure 11A. The predictive model was able to accurately classify healthy samples with 97.9% (representing specificity), lung cancer with 100%, and breast cancer with 92.9% regardless of sample dilution (Figure 11B). More breast cancer samples may be added to ensure sufficient training and improve predictive accuracy.

[0081] In one embodiment, system 100 and / or its components or subsystems may include computing devices, microprocessors, modules, and other computers or computing devices, which may be any programmable devices that accept digital data as input, are configured to process the input according to instructions or algorithms, and provide the results as output. In one embodiment, the computing devices and other such devices described herein may be, include, contain, or be coupled with a central processing unit (CPU) configured to execute instructions of a computer program. Thus, the computing devices and other such devices described herein are configured to perform basic arithmetic, logical, and input / output operations.

[0082] The computing devices and other devices described herein may include memory. Memory may include volatile or non-volatile memory as required by the coupled computing device or processor, not only to provide space for executing instructions or algorithms, but also to provide space for storing the instructions themselves. In one embodiment, volatile memory may include, for example, random-access memory (RAM), dynamic random-access memory (DRAM), or static random-access memory (SRAM). In one embodiment, non-volatile memory may include, for example, read-only memory, flash memory, ferroelectric RAM, hard disk, floppy disk, magnetic tape, or optical disk storage. These embodiments are given merely as examples and are not intended to limit the scope of this disclosure; therefore, the foregoing list is by no means limited to the types of memory that may be used.

[0083] In one embodiment, a system or its components may comprise or include various modules or engines, each of which is constructed, programmed, configured, or possibly adapted to autonomously perform a function or set of functions. As used herein, the term “engine” is defined, for example, as an arrangement of components implemented using hardware such as an existing device, component, or application-specific integrated circuit (ASIC) or field-programmable gate array (FPGA), or as a combination of hardware and software such as a microprocessor system and a set of program instructions, which adapt the engine to perform a specific function that transforms the microprocessor system into a dedicated device (while being executed). An engine may also be implemented as a combination of two functions: some functions facilitated solely by hardware, and others facilitated by a combination of hardware and software. In some implementations, at least part, and possibly all, of the engine may run on the processors of one or more computing platforms, which consist of hardware running operating systems, system programs, and application programs (e.g., one or more processors, data storage devices such as memory or drive memory, input / output equipment such as network interface devices, video devices, keyboards, mice or touchscreen devices, etc.), while also implementing engines that utilize multitasking, multithreading, distributed processing (e.g., clusters, peer-to-peer, cloud, etc.) or other such techniques, as needed. Accordingly, each engine may be implemented in a variety of physically feasible configurations and should not generally be limited to any particular implementation exemplified herein unless such limitations are explicitly stated. In addition, an engine may itself consist of two or more sub-engines, each of which may be considered an engine in itself.Furthermore, while in the embodiments described herein each of the various engines corresponds to a defined autonomous function, it should be understood that in other embodiments considered, each function may be distributed among two or more engines. Similarly, in other embodiments considered, multiple defined functions may be implemented by a single engine that performs those functions possibly in parallel with other functions, or they may be distributed among a different set of engines than those specifically shown in the examples herein.

[0084] Various embodiments of systems, devices, and methods have been described herein. These embodiments are given merely as examples and are not intended to limit the scope of the claimed invention. Furthermore, it should be understood that various features of the described embodiments can be combined in various ways to create numerous additional embodiments. Moreover, various materials, dimensions, shapes, configurations, and locations, etc., have been described for use with the disclosed embodiments, but others beyond those disclosed may be used without exceeding the scope of the claimed invention.

[0085] Those skilled in the art will recognize that the subject matter of this specification may contain fewer features than those shown in any of the individual embodiments described above. The embodiments described herein are not intended to be a comprehensive presentation of how the various features of the subject matter of this specification can be combined. Accordingly, the embodiments are not mutually exclusive combinations of features, but rather, as those skilled in the art will understand, various embodiments may include different combinations of individual features selected from different individual embodiments. Moreover, elements described in relation to one embodiment may be implemented in other embodiments even when not described in that embodiment, unless otherwise noted.

[0086] A dependent claim may refer to a specific combination of one or more other claims within the scope of the claims, but other embodiments may also include combinations of a dependent claim with the subject matter of each other dependent claim, or combinations of one or more features with other dependent or independent claims. Such combinations are proposed herein unless otherwise stated that a particular combination is not intended.

[0087] Any incorporation by reference to the above documents is limited so as not to include any subject matter that contradicts the express disclosure herein. Any incorporation by reference to the above documents is further limited so as not to include any claims not contained in those documents by reference herein. Any incorporation by reference to the above documents is further limited so as not to include any definitions provided in those documents by reference herein unless they are expressly included herein.

[0088] It is explicitly intended that, in order to interpret the scope of the claims, the provisions of Section 112(f) of the U.S. Patent Act will not be enforced unless the specific terms “means for” or “steps for” are included in the claims. [Explanation of symbols]

[0089] 100 flow paths, system 102 whole blood samples 104 Centrifugal Separation 106 Serum 108 Spectrophotometer, light sensor 110 Spectroscopic Signature 200 First System 300 Second System 400 micro cuvettes 410 Blood coagulation chamber 412 Self-locking cap 414 Coagulation Strainer 416 Serum channel plug 420 Serum Analysis Chamber 422 Container 424 Drainage Channel 500 flow paths 502 serum samples 504 Melt 506 base samples 508 First revised sample 510 Second revised sample 512 Attenuated Total Reflectance ("ATR") Crystal 514 Infrared ("IR") Reflection Sampling Card 516 Fourier Transform Infrared Spectroscopy (FTIR) 600 Calculation Engines 602 Computational Algorithms 604 Data Storage Devices 606 Machine Learning Parameters 608 Machine Learning Models 610 Interface 612 Input data 614 results 616 Feedback

Claims

1. A method of diagnosing a disease, A step of analyzing a sample by absorption spectroscopy in the near-infrared and / or mid-infrared range to create a spectral signature, wherein the sample is obtained from a patient, and The steps include receiving the spectral signature with a processor, The steps include: determining by the processor whether the spectral signature consequently indicates the presence of the disease; The steps include outputting the above result using the processor and Methods that include...

2. The method according to claim 1, further comprising the step of providing the patient with a treatment to treat the disease when the result indicates that the disease is present.

3. The method according to claim 1, wherein the sample is a lysed sample.

4. The method according to claim 3, wherein the lysed sample is obtained by adding a solubilizing solution or a homogenizing solution to a serum sample.

5. The method according to claim 4, wherein the solubilizing solution or the homogenizing solution is added to the serum sample in a 1:1 ratio.

6. The method according to claim 3, wherein at least one of an optical molecule binding solution or a protein degradation reagent is added to the lysate sample.

7. The method according to claim 1, further comprising the step of drying the sample on an IR reflectance sampling card or a non-IR absorption sampling card.

8. The method according to claim 7, wherein the IR reflection sampling card is coated with aluminum.

9. The step of the processor determining whether the spectral signature consequently indicates the presence of the disease is: The steps include providing the spectral signature to a computation engine that includes a model architecture and one or more model parameters, The steps include: executing a computation algorithm configured to provide the results based on the spectral signature, the model architecture, and one or more model parameters using the computation engine; The method according to claim 1, including the method described in claim 1.

10. A step of providing feedback to the calculation engine indicating the accuracy of the above results, The steps include updating one or more model parameters based on the above results, the spectral signature, and the feedback. The method according to claim 9, further comprising:

11. It is a system for diagnosing the presence of a disease. Memory and It is a processor, Receiving a spectral signature created by absorption spectroscopy in the near-infrared and / or mid-infrared range of a patient's sample, The spectral signature determines whether it consequently indicates the presence of the disease, Output the above result and To do this, a processor configured to execute instructions stored in the memory and A system equipped with these features.

12. The system according to claim 11, wherein the sample is a serum sample.

13. The system according to claim 11, wherein the sample is a whole blood sample.

14. The system according to claim 11, wherein the sample is a lysed sample.

15. The system according to claim 13, wherein the lysed sample is obtained by adding a solubilizing solution or a homogenizing solution to the serum sample.

16. The system according to claim 14, wherein the solubilizing solution or the homogenizing solution is added to the serum sample in a 1:1 ratio.

17. The processor determines whether the spectral signature indicates the presence of the disease. To provide the spectral signature to a computation engine that includes a model architecture and one or more model parameters, The computational engine executes a computational algorithm configured to provide the results based on the spectral signature, the model architecture, and one or more model parameters. The system according to claim 11, configured to be determined by

18. The aforementioned processor, To provide the calculation engine with feedback indicating the accuracy of the results, Based on the above results, the spectral signature, and the feedback, update one or more model parameters. The system according to claim 17, further configured to execute instructions in the memory in order to do so.

19. A method for detecting pathogens in a sample, The steps include receiving the patient's whole blood sample into a coagulation cuvette, The steps include operating the coagulation cuvette to release a serum sample into the analysis chamber, The steps include inserting the cuvette into a spectrophotometer that uses near-infrared and / or mid-infrared spectra, A step of determining whether the pathogen is present in the serum sample, A step of outputting a result indicating whether the aforementioned pathogen was detected. Methods that include...

20. There is a system for detecting pathogens in a sample, Coagulation cuvette and, A spectrophotometer that uses near-infrared and / or mid-infrared spectra, A processor associated with the spectrophotometer and configured to perform analysis of the spectrophotometric signature output by the spectrophotometer. A system equipped with these features.

21. A serum isolation cuvette, A serum analysis chamber including a sample container and one or more serum channels on the sample container, A coagulation chamber comprising a coagulant for coagulating a sample introduced into the coagulation chamber, a coagulation strainer for removing all coagulated cells from the sample, and a channel plug. Equipped with, A serum separation cuvette wherein the channel plug and the serum channel form a releaseable seal between the coagulation chamber and the analysis chamber, allowing sample serum to flow into the analysis chamber.

Images