Systems and methods for disease prediction

Abstract

A system (100) and a method (200) for predicting the presence of a disease in a subject are disclosed. In one aspect, the system (100) includes one or more processing units (101), a medical database (112) coupled to the one or more processing units (101), and a disease prediction module (110). The module (110) is configured to receive a plurality of parameters associated with the subject. Furthermore, the module (110) is configured to determine a threshold value for a sensitivity and / or specificity associated with the trained machine learning model. Furthermore, the module (110) is configured to use the trained machine learning model to predict the presence of a disease in the subject based on a desired sensitivity and specificity and the plurality of parameters associated with the subject, and output the prediction to an output unit (105).

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority or benefit under 35 U.S.C. Section 119 of U.S. Provisional Patent Application No. 63 / 268,435, filed February 24, 2022, Indian Patent Application No. 202231010035, filed February 24, 2022, and Indian Patent Application No. 202331002562, filed January 12, 2023, all of which are incorporated by reference herein in their entireties.

[0002] The present disclosure relates generally to the fields of medicine, subject health, and analysis of samples, and more particularly to the field of predicting the presence of disease in a subject. [Background technology]

[0003] Disease prediction is important not only to ensure timely treatment to patients but also to manage the resources and logistics involved in the disease management process. For example, the pandemic caused by the SARS-CoV-2 virus has increased the burden on healthcare systems worldwide. The accelerated increase in the number of individuals who need to be tested for the detection of the SARS-CoV-2 virus has resulted in shortages of infectious disease testing kits such as reverse transcriptase polymerase chain reaction (RT-PCR) testing kits, delays in providing test results to individuals, and delays in providing timely medical assistance to patients when the presence of the SARS-CoV-2 virus was not suspected. Prediction of the presence of one or more pathogens, such as viruses, e.g., SARS-CoV-2, bacteria, etc., in patients can be important in such scenarios. Summary of the Invention [Problem to be solved by the invention]

[0004] Currently, there is no reliable method that can efficiently predict the presence of a pathogen, such as a virus, e.g., SARS-CoV-2 virus, in a specimen without RT-PCR or antigen testing. There is a continuing need for improved systems and methods to predict the presence of a pathogen, such as a virus or bacteria, in a specimen and / or detect increased infection rates in a population to improve utilization of medical resources. [Means for solving the problem]

[0005] A system for predicting the presence of a disease in a subject is disclosed. In one embodiment of the present disclosure, the system includes a processing unit and a memory. Further, the memory includes a disease prediction module configured to receive a plurality of parameters associated with the subject and determine a threshold value for sensitivity and / or specificity associated with a trained machine learning model. Further, the module is configured to predict the presence of the disease in the subject and the plurality of parameters associated with the subject using the trained machine learning model, and output the prediction to an output unit.

[0006] In another aspect of the present disclosure, a method for predicting the presence of a disease in a subject is disclosed. The method includes receiving a plurality of parameters associated with the subject. Further, the method includes determining a threshold value for a sensitivity and / or specificity associated with the trained machine learning model. Further, the method includes predicting the presence of the disease in the subject based on the sensitivity and specificity of the disease and the plurality of parameters associated with the subject using the trained machine learning model. Further, the method includes outputting the prediction to an output unit.

[0007] In an embodiment, the present disclosure includes an article of manufacture including a non-transitory computer-readable medium having instructions encoded thereon, the instructions configured to cause one or more processors to perform a method including predicting the presence of a disease in accordance with the present disclosure.

[0008] In an embodiment, the present disclosure includes a non-transitory computer-readable medium having instructions encoded thereon, the instructions configured to cause one or more processors to execute a method including predicting the presence of a disease in accordance with the present disclosure.

[0009] In embodiments, the present disclosure includes predicting the presence of disease in a subject according to the present disclosure and subsequently treating the subject in need thereof.

[0010] In an embodiment, the present disclosure includes a system for predicting the presence of a disease in a population. In one aspect of the present disclosure, the system includes a processing unit and a memory. Further, the memory includes a disease prediction module configured to receive a plurality of parameters associated with one or more subjects and determine a threshold value for sensitivity and / or specificity associated with the trained machine learning model. Further, the module is configured to use the trained machine learning model to predict the presence of a disease in one or more subjects or a population of subjects, and the plurality of parameters associated with the subject or population of subjects, and output the prediction to an output unit.

[0011] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the following Description. It is not intended to identify feature or essential features of the claimed subject matter. Moreover, the claimed subject matter is not limited to implementations that solve any or all of the disadvantages noted in any part of this disclosure.

[0012] The invention is further described below with reference to exemplary embodiments illustrated in the accompanying drawings. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 illustrates a system for predicting the presence of a disease in a subject, according to one embodiment. [Diagram 2]FIG. 1 illustrates a process flow for predicting the presence of disease in a subject, according to one embodiment of the present disclosure. [Diagram 3] FIG. 1 illustrates a receiver operating characteristic (ROC) curve for selecting thresholds for sensitivity and specificity associated with a trained machine learning model for a requirements scenario, according to one embodiment of the present disclosure. [Figure 4] FIG. 1 depicts the ROC curve for a 48-parameter ensemble model computed using the test set of the present disclosure. [Diagram 5] Model interpretability: A diagram depicting features arranged in order of importance as computed by the MIMIC LightGBM explainer. [Figure 6A] FIG. 1 depicts use case embodiments such as a pandemic scenario and an endemic scenario use case in region (1) of the ROC curve. [Figure 6B] FIG. 1 depicts use case embodiments such as a pandemic scenario and an endemic scenario use case in region (2) of the ROC curve. [Figure 6C] FIG. 1 depicts use case embodiments such as a pandemic scenario and an endemic scenario use case in region (3) of the ROC curve. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, systems, methods, and articles of manufacture for carrying out the embodiments of the present disclosure will be described in detail. Various embodiments are described with reference to the drawings, and like reference numerals are used to refer to like elements throughout. In the following description, for purposes of explanation, numerous specific details are set forth to provide a thorough understanding of one or more embodiments. It will be apparent that such embodiments may be practiced without these specific details. In other instances, well-known materials or methods have not been described in detail to avoid unnecessarily obscuring the embodiments of the present disclosure. While the present disclosure is susceptible to the effects of various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and described in detail herein. However, it is to be understood that there is no intention to limit the present disclosure to the particular forms disclosed, but rather, the present disclosure is intended to encompass all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure. The disclosed embodiments provide a system and method for predicting disease in a subject.

[0015] Advantages of embodiments of the present disclosure include, among others, superior and robust prediction of disease caused by one or more pathogens to ensure timely treatment to patients or subjects in need thereof, as well as improved management of resources and logistics involved in the disease management process. Embodiments of the present disclosure reduce the burden on healthcare systems around the world, prevent shortages of test kits, reduce waiting times for individuals' test results, and reduce waiting times for medical assistance to patients when the presence of a pathogen is not suspected.

[0016] While COVID-19 testing using RT-PCR (reverse transcriptase polymerase chain reaction) and antigen testing has been massively expanded since the outbreak of the COVID-19 pandemic, there are multiple scenarios in which CBC / Diff-based predictive algorithms can be applied to the clinical value chain in both pandemic and endemic scenarios. First, CBC / Diff-based predictive algorithms can serve as a substitute for testing in countries dealing with the unavailability or shortage of such tests. Second, rapid flagging by highly sensitive hematology-based machine learning (ML) algorithms (e.g., see Algorithm A described below with a threshold of 0.41) can reduce the testing burden in pandemic scenarios. Third, algorithms with high negative predictive value (NPV) (e.g., see Algorithm A with a threshold of 0.33) can reduce the number of patients who are tested prior to procedures such as surgery, thereby shortening the time to patient care. Fourth, in endemic scenarios, algorithms with high specificity (e.g., see Algorithm A with a threshold of 0.67) can be used to flag asymptomatic individuals who can be tested to confirm infection. Thus, the same ML algorithm, such as Algorithm A, with different classification thresholds (e.g., thresholds between 0.4 and 1.0, between 0.4 and 0.97, between 0.4 and 0.95, between 0.4 and 0.7, between 0.4 and 0.6, between 0.4 and 0.5, or between 0.4 and 0.45, between 0.80 and 0.95, between 0.99 and 1.0, etc.) can be used in multiple scenarios described herein.

[0017] Some definitions The following terms utilized in accordance with the present disclosure shall be understood to have the following meanings, unless otherwise indicated: unless otherwise clear from context, (i) the term "a" may be understood to mean "at least one," (ii) the terms "comprising" and "including" may be understood to encompass the itemized element or step, whether presented by itself or with one or more additional elements or steps, and (iii) when ranges are provided, the endpoints are included.

[0018] Use of the term "at least one" is understood to include any quantity greater than one, including but not limited to 1, as well as 2, 3, 4, 5, 10, 15, 20, 30, 40, 50, 100, etc. The term "at least one" can extend to 100 or 1000 or more, depending on the term with which it is accompanied, and in addition, the quantity of 100 / 1000 should not be considered limiting, as higher limits may also provide satisfactory results. Additionally, use of the term "at least one of X, Y, and Z" is understood to include X only, Y only, and Z only, as well as any combination of X, Y, and Z. Use of ordinal terms (i.e., "first," "second," "third," "fourth," etc.) is merely to distinguish between two or more items and does not imply, for example, any order or sequence or importance to one item compared to another item, or any additional order.

[0019] Use of the term "or" in the claims is used to mean inclusive "and / or" unless expressly indicated to refer to alternatives only or unless the alternatives are mutually exclusive. For example, a condition "A or B" is satisfied by any of the following: A is true (or exists) and B is false (or does not exist), A is false (or does not exist) and B is true (or exists), and both A and B are true (or exist).

[0020] As used herein, any reference to "one embodiment," "embodiment," "some embodiments," "an example," "for example," or "an example" means that a particular element, feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment. The appearances of the phrases "in some embodiments" or "an example" in various places in the specification, for example, do not necessarily all refer to the same embodiment. Moreover, all references to one or more embodiments or examples should be construed as non-limiting on the scope of the claims.

[0021] Throughout this application, the term "about" is used to indicate that a value includes the inherent variation of error for a configuration / apparatus / device, the method used to determine the value, or the variation that exists among study subjects. For example, and without limitation, when the term "about" is utilized, the specified value may vary plus or minus 20%, or 15%, or 12%, or 11%, or 10%, or 9 percent, or 8 percent, or 7 percent, or 6 percent, or 5 percent, or 4 percent, or 3 percent, or 2 percent, or 1 percent from the specified value, such variations being appropriate to carry out the disclosed method and understood by those of skill in the art. In embodiments, the term "about" when used herein with reference to a value refers to a value similar to the value referenced in the context. In general, a person of skill in the art familiar with the context will understand the relevant degree of variation encompassed by "about" in that context. For example, in some embodiments, the term "about" can encompass values ​​within a range of 25%, 20%, 19%, 18%, 17%, 16%, 15%, 14%, 13%, 12%, 11%, 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, or less of the stated value.

[0022] The term "or combinations thereof" as used herein refers to all permutations and combinations of the listed items preceding the term. For example, "A, B, C, or combinations thereof" is intended to include at least one of A, B, C, AB, AC, BC, or ABC, and also includes BA, CA, CB, CBA, BCA, ACB, BAC, or CAB if order is important in a particular context. Continuing with this example, combinations including repetitions of one or more items or terms, such as BB, AAA, AAB, BBC, AAABCCCC, CBBAAA, CABABB, etc., are expressly included. Those skilled in the art will understand that there is generally no limit to the number of items or terms in any combination unless otherwise clear from the context.

[0023] As used herein, the term "CBC / Diff" or "Blood Count with Differential" refers to a measure of the number of red blood cells, white blood cells, and platelets in the blood, including different types of white blood cells (neutrophils, lymphocytes, monocytes, basophils, and eosinophils). In embodiments, the amount of hemoglobin (the substance in blood that carries oxygen) and hematocrit (the amount of whole blood that is made up of red blood cells) are also measured.

[0024] As used herein, a "diagnostic test" is a step or series of steps taken or performed to determine whether a patient has a disease, disorder, or condition, and / or to obtain information useful in classifying a disease, disorder, or condition into a phenotypic category or any category that has important implications for the prognosis of the disease, disorder, or condition, or the likelihood of response to treatment of the disease, disorder, or condition (either treatment in general or any specific treatment). Similarly, "diagnosis" refers to providing any type of diagnostic information, including, but not limited to, whether a subject has or is likely to develop a disease, disorder, or condition, the state, stage, or characteristics of the disease, disorder, or condition present in the subject, information about the nature or classification of the tumor, information about the prognosis, and / or information useful in selecting an appropriate treatment or additional diagnostic tests. Selection of treatment may include selection of a particular therapeutic agent or other treatment such as surgery, radiation, selection of whether to withhold or perform treatment, selection of a dosing regimen (e.g., frequency or level of one or more doses of a particular therapeutic agent or combination of therapeutic agents), etc. The selection of additional diagnostic tests may include more specific tests for a given disease, disorder, or condition.

[0025] As used herein, the terms "disease" and "disorder" are used interchangeably to refer to a subject's illness, including a harmful deviation from the normal structural or functional state of an organism.Non-limiting examples of disease / disorder include a subject having one or more viral infections, one or more bacterial infections, one or more fungal or parasitic infections, or sepsis.

[0026] The term "viral infection" as used herein refers to the entry, proliferation, and / or presence of a virus in a cell or subject. In one embodiment, the viral infection is an "active" infection, i.e., an infection in which the virus is replicating in a cell or subject. Such an infection is characterized by the spread of the virus from a cell, tissue, and / or organ that is initially infected by the virus to other cells, tissues, and / or organs. The infection can also be a latent infection, i.e., an infection in which the virus is not replicating.

[0027] As used herein, the term "bacterial infection" refers to the invasion, proliferation, and / or presence of bacteria within a cell or subject.

[0028] As used herein, the term "fungal infection" refers to the invasion, growth, and / or presence of a fungus within a cell or subject.

[0029] As used herein, the term "pathogen infection" refers to the invasion, proliferation, and / or presence of a pathogen (such as a bacterial, fungal, parasitic, or viral pathogen) within a cell or subject.

[0030] The term "sample" as used herein refers to a biological sample obtained or derived from a human subject, as described herein. In some embodiments, the biological sample comprises a biological tissue or bodily fluid. In some embodiments, the biological sample may comprise blood, blood cells, tissue or fine needle biopsy samples, cell-containing bodily fluids, free floating nucleic acids, cerebrospinal fluid, lymphatic fluid, tissue biopsy specimens, surgical specimens, other bodily fluids, secretions, and / or excretions, and / or cells therefrom. In some embodiments, the biological sample comprises cells obtained from an individual, e.g., a human or animal subject. In some embodiments, the obtained cells are or include cells from the individual from whom the sample was obtained. In some embodiments, the sample is a "primary sample" obtained directly from the source of interest by any suitable means. For example, in some embodiments, the primary biological sample is obtained by a method selected from the group consisting of biopsy (e.g., fine needle aspiration or tissue biopsy), surgery, collection of bodily fluids (e.g., blood). In some embodiments, the sample is cardiac tissue obtained from the subject. In some embodiments, as will be clear from the context, the term "sample" refers to a preparation obtained by processing a primary sample (e.g., by removing one or more components and / or by adding one or more agents). For example, filtration using a semi-permeable membrane. Another example of sample processing may be a plasma sample treated with an anticoagulant selected from the group consisting of EDTA, heparin, and citrate. Another example of sample processing may be a sample that is processed to isolate one or more proteins (e.g., by capturing the proteins with one or more antibodies). A "processed sample" may include, for example, nucleic acids or polypeptides extracted from a sample or obtained by subjecting a primary sample to techniques such as amplification or reverse transcription of mRNA, isolation and / or purification of specific components.

[0031] As used herein, the term "SARS CoV-2 variant" or "viral variant" or simply "variant", when used in reference to SARS CoV-2 (as the context makes clear), is a SARS CoV-2 virus that contains one or more genetic changes compared to a SARS CoV-2 reference strain or one or more predominant viral variants already circulating in a population. In some embodiments, the SARS CoV-2 variant contains one or more genomic variants in the S gene sequence. In some embodiments, the SARS CoV-2 variant contains one or more genomic variants relative to a reference genome sequence or a portion thereof. In some embodiments, the SARS CoV-2 variant contains one or more genomic variants in the S gene sequence compared to the S gene sequence of the reference genome. In some embodiments, the reference genome sequence corresponds to that of the Wuhan-Hu1 strain (the first genetic sequence identified) or the USA-WA1 / 2020 strain (first identified in the United States) or a portion thereof. In some embodiments, the SARS CoV-2 variant contains one or more genomic variants compared to one or more predominant viral variants circulating in a population. In some embodiments, the SARS CoV-2 variant comprises one or more genomic variants in the S gene sequence compared to the S gene sequence of the predominant viral variant circulating in the population.

[0032] The term "subject" as used herein refers to an organism, e.g., a mammal (e.g., a human). In some embodiments, the human subject is an adult subject, an adolescent subject, or a pediatric subject. In some embodiments, the subject is at least 50 years old, at least 55 years old, at least 60 years old, at least 65 years old, at least 70 years old, at least 75 years old, or at least 80 years old. In some embodiments, the subject suffers from a disease, disorder, or illness, e.g., a disease, disorder, or illness that can be treated as provided herein. In some embodiments, the subject is susceptible to a disease, disorder, or illness, and in some embodiments, a susceptible subject is predisposed to developing a disease, disorder, or illness and / or exhibits an increased risk of developing a disease, disorder, or illness (compared to the average risk observed in a reference subject or population). In some embodiments, the subject exhibits one or more symptoms of a disease, disorder, or illness. In some embodiments, the subject does not exhibit a particular symptom (e.g., clinical symptoms of a disease) or characteristic of a disease, disorder, or illness. In some embodiments, the subject does not exhibit any symptoms or characteristics of a disease, disorder, or illness. In some embodiments, the subject is a patient. In some embodiments, the subject is an individual who is and / or has been diagnosed and / or treated. In some embodiments, the human subject is an adult subject, an adolescent subject, or a pediatric subject. In some embodiments, the subject is at risk for SARS CoV-2 viral infection. In some embodiments, the subject is at risk for viral infection, bacterial infection, sepsis, or a combination thereof. In some embodiments, the subject has or is suspected of having COVID-19 infection. In some embodiments, the subject is susceptible to COVID-19 infection, and in some embodiments, the susceptible subject is predisposed to developing COVID-19 and / or exhibits an increased risk of developing COVID-19 (compared to the average risk observed in a reference subject or population). In some embodiments, the subject exhibits one or more symptoms of SARS CoV-2 infection.In some embodiments, the subject does not exhibit any symptoms or characteristics of COVID-19. In some embodiments, the subject does not exhibit any symptoms or characteristics of COVID-19. In some embodiments, the subject is a patient. In some embodiments, the subject is an individual who is and / or has been diagnosed and / or treated.

[0033] The term "substantially" as used herein means that the subsequently described event or circumstance occurs completely, or that the subsequently described event or circumstance occurs to a significant extent. For example, when associated with a particular event or circumstance, the term "substantially" means that the subsequently described event or circumstance occurs at least 80% of the time, or at least 85% of the time, or at least 90% of the time, or at least 95% of the time. The term "substantially adjacent" can mean that two items are 100% adjacent to each other, or that two items are close to each other but not 100% adjacent to each other, or that a portion of one of two items is close to the other item but not 100% adjacent to the other item.

[0034] The term "threshold" as used herein refers to a value (or values) that is used as a reference to obtain information about and / or classify a measurement result, e.g., a measurement result obtained in an analysis. The threshold can be determined based on one or more control samples. The threshold can be determined before, at the same time, or after the measurement of interest is performed. In some embodiments, the threshold can be a range of values. In some embodiments, the threshold can be a value (or range of values) reported in the relevant field (e.g., values ​​found in standard tables).

[0035] The term "treatment" or "treating" of a subject includes the application or administration of a compound to a subject for the purpose of delaying, stalling, stabilizing, curing, curing, mitigating, alleviating, altering, correcting, making less worse, improving, ameliorating, or affecting a disease or condition, a symptom of a disease or condition, or the risk (or susceptibility) of a disease or condition. The term "treating" refers to any indicator of success in treating or improving an injury, pathology, or condition, including any objective or subjective parameter, such as palliating, remission, reducing the rate of deterioration, reducing the severity of a disease, stabilizing, attenuating symptoms, or making the injury, pathology, or condition more tolerable to the subject, slowing the rate of degeneration or decline, making the end point of degeneration less debilitating, or improving the physical or mental health of the subject. In embodiments, the term "treating" refers to reducing or ameliorating the progression, severity, and / or duration of COVID-19 caused by the administration of one or more therapies (e.g., one or more therapeutic agents), or improving one or more symptoms of COVID-19. In certain embodiments, the term "treat" refers to improving a measurable physical parameter of COVID-19. In embodiments, the term "treat" refers to reducing or improving the progression, severity, and / or duration of a viral, bacterial, or fungal infection caused by administration of one or more therapies (e.g., one or more therapeutic agents), or improving one or more symptoms of a viral, bacterial, or fungal infection. In certain embodiments, the term "treatment" refers to improving a measurable physical parameter of a viral, bacterial, or fungal infection. In embodiments, the term "treat" refers to reducing or improving the progression, severity, and / or duration of sepsis caused by administration of one or more therapies (e.g., one or more therapeutic agents), or improving one or more symptoms of sepsis. In certain embodiments, the term "treat" refers to improving a measurable physical parameter of sepsis.In embodiments, "treating" refers to altering the natural or present state of a subject.

[0036] The term "therapeutically effective amount" as used herein means an amount of a compound that, when administered to a subject to treat or prevent a particular disorder, disease, or condition, is sufficient to effect such treatment or prevention of the disorder, disease, or condition. Dosage and therapeutically effective amounts may vary depending on a variety of factors, including, for example, the activity of the particular agent used, the age, weight, general health, sex, and diet of the subject, the time of administration, the route of administration, the rate of excretion, and any drug combinations, if applicable, the effect the practitioner desires the compound to have on the subject and the properties of the compound (e.g., bioavailability, stability, efficacy, toxicity, etc.), and the particular disorder from which the subject suffers. Additionally, a therapeutically effective amount administered intravenously may depend on the subject's blood parameters, such as lipid profile, insulin levels, blood glucose, or liver metabolism. Therapeutically effective amounts also vary depending on the disease state, organ function, or severity of the underlying disease or comorbidities. Such appropriate dosages may be determined using any available assay. When one or more compounds or therapeutic agents are to be administered to a human, a physician may, for example, prescribe a relatively low dose initially, subsequently increasing the dose until an appropriate response is obtained.

[0037] As used herein, the term "viral disease" refers to a pathological condition resulting from the presence of a virus in a cell or subject, or the invasion of a cell or subject by a virus.

[0038] As used herein, the term "influenza virus disease" refers to a pathological condition resulting from the presence of an influenza virus (e.g., influenza A virus or B virus) in a cell or subject, or the invasion of a cell or subject by an influenza virus. In certain embodiments, the term refers to a respiratory disease caused by an influenza virus.

[0039] As used herein, the term "COVID-19 disease" refers to a pathological condition resulting from the presence of SARS-CoV-2 (e.g., SARS-CoV-2 variants) virus in a cell or subject or invasion of a cell or subject by the SARS-CoV-2 virus or its variants. In certain embodiments, the term refers to a respiratory disease caused by a coronavirus, such as SARS-CoV-2 or its variants.

[0040] The term "sepsis" has been used to describe a variety of clinical conditions associated with systemic manifestations of inflammation accompanied by infection. Because of clinical similarities with inflammatory responses secondary to non-infectious etiologies, identification of sepsis has been a particularly challenging diagnostic problem.Sepsis definitions may change over time and across different clinical / hospital systems (e.g., Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, Bellomo R, Bernard GR, Chiche JD, Coopersmith CM, Hotchkiss RS. The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA. 2016 Feb 23;315(8):801-10. Bone RC, Balk RA, Cerra FB, et al. American College of Chest Physicians / Society of Critical Care Medicine Consensus Conference: definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med. 1992;20(6):864-874. Levy MM, Fink MP, Marshall JC, et al. International Sepsis Definitions Conference. 2001; SCCM / ESICM / ACCP / ATS / SIS International Sepsis Definitions Conference. Intensive Care Med. 2003, 29(4):530-538. Centers for Disease Control and Prevention. Hospital toolkit for adult sepsis surveillance. 2018 (see, e.g., www.cdc.gov / sepsis / pdfs / Sepsis-Surveillance-Toolkit-Mar-2018_508.pdf). All such definitions of sepsis, including but not limited to the examples cited above, are intended for purposes of this application.

[0041] DETAILED DESCRIPTION OF CERTAIN EMBODIMENTS 1 is a block diagram of a system 100 in which an embodiment may be implemented, for example, as a system 100 for predicting the presence of a disease, configured to perform the processes described therein. In FIG. 1, said system 100 includes a processing unit 101, a memory 102, a storage unit 103, an input unit 104, a bus 106, an output unit 105, and a network interface 107.

[0042] As used herein, processing unit 101 refers to any type of computational circuitry, such as, but not limited to, a microprocessor, a microcontroller, a complex instruction set computing microprocessor, a reduced instruction set computing microprocessor, a very long instruction word microprocessor, an explicitly parallel instruction computing microprocessor, a graphics processor, a digital signal processor, or any other type of processing circuitry. Processing unit 101 may also include embedded controllers, such as general purpose or programmable logic devices or arrays, application specific integrated circuits, single chip computers, and the like.

[0043] The memory 102 may be a volatile memory and a non-volatile memory. The memory 102 may be coupled to communicate with the processing unit 101. The processing unit 101 may execute instructions and / or codes stored in the memory 102. A variety of computer readable storage media may be stored in and accessed from the memory 102. The memory 102 may include any suitable element for storing data and machine readable instructions, such as read only memory, random access memory, erasable programmable read only memory, electrically erasable programmable read only memory, a hard drive, a removable media drive for handling compact discs, digital video discs, diskettes, magnetic tape cartridges, memory cards, and the like. In this embodiment, the memory 102 includes a disease prediction module 110 stored in the form of machine readable instructions in any of the above-mentioned storage media, and may be in communication with and executed by the processor 101. When executed by the processor 101, the disease prediction module 110 causes the processor 101 to predict the presence of a disease in a subject. The method steps performed by the processor 101 to implement the above-described functions are detailed in FIG.

[0044] The storage unit 103 may be a non-transitory storage medium that stores the medical database 112. The medical database 112 is a repository of patient data including blood parameters maintained by a healthcare service provider. The input unit 104 may include input means such as a keypad capable of receiving an input signal, a touch-sensitive display, a camera (such as a camera that receives gesture-based input), etc. The bus 106 serves as an interconnection between the processing unit 101, the memory 102, the storage unit 103, the input unit 104, the output unit 105, and the network interface 107.

[0045] Those skilled in the art will appreciate that the hardware depicted in Figure 1 may vary for a particular implementation. For example, other peripheral devices such as local area network (LAN) / wide area network (WAN) / wireless (e.g., Wi-Fi) adapters, graphics adapters, disk controllers, input / output (I / O) adapters, etc., such as optical disk drives, may also be used in addition to or in place of the depicted hardware. The depicted examples are provided for illustrative purposes only and are not intended to imply architectural limitations with respect to the present disclosure.

[0046] A system 100 according to one embodiment of the present disclosure includes an operating system that uses a graphical user interface. The operating system allows multiple display windows to be presented simultaneously within the graphical user interface, each display window providing an interface to a different application or a different instance of the same application. A cursor within the graphical user interface may be manipulated by a user via a pointing device. The position of the cursor may be changed and / or an event, such as clicking a mouse button, may be generated to actuate a desired response. If appropriately modified, one of a variety of commercial operating systems may be used, such as a version of Microsoft Windows™, a product of Microsoft Corporation, Redmond, Washington. The operating system may be modified or created according to the present disclosure as described.

[0047] Further provided is a non-transitory computer readable medium including executable instructions that, when executed, cause a processor to perform operations including the methods provided herein. In an embodiment, the present disclosure includes an article of manufacture, such as a system or a component thereof, including a non-transitory computer readable medium having instructions encoded thereon, the instructions configured to cause one or more processors to perform a method including predicting the presence of a disease in accordance with the present disclosure. In an embodiment, the present disclosure includes a non-transitory computer readable medium including instructions encoded thereon, the instructions configured to cause one or more processors to perform a method including predicting the presence of a disease in accordance with the present disclosure.

[0048] FIG. 2 shows a process flow 200 for predicting the presence of disease in a subject according to another embodiment. In step 201, a plurality of parameters associated with the subject are received. The parameters associated with the subject may be obtained from a complete blood count / differential test performed on the subject. In particular, the parameters may include red blood cell (RBC) parameters, white blood cell (WBC) parameters, and platelet parameters. The RBC parameters include hemoglobin level, hematocrit, RBC size, hemoglobin level (normal, high, low) in the individual's RBCs, mean corpuscular volume, mean corpuscular hemoglobin concentration, normal RBC, number of RBCs with hemoglobin concentration ≧28 g / dL and ≦41 g / dL, number of RBCs with hemoglobin volume ≧60 fL and ≦120 fL, hemoglobin distribution width, RBC mean corpuscular volume, RBC volume distribution width, total RBC count, hemoglobin content distribution width, and hemoglobin content average. RBC size-based parameters include RBC counts below 60 fL and above 120 fL. Additional parameters such as average cellular hemoglobin concentration, number of RBCs with hemoglobin concentration above 41 g / dL, and number of RBCs with hemoglobin concentration below 28 g / dL are also considered. Subject-associated parameters may be obtained from a complete blood count / differential test performed on the subject. In particular, the parameters may include red blood cell (RBC) parameters, white blood cell (WBC) parameters, and platelet parameters. In an embodiment, the RBC parameters are preselected and include one or more of hemoglobin level, hematocrit, RBC size, hemoglobin level (normal, high, low) in the individual's RBCs, mean corpuscular volume, mean corpuscular hemoglobin concentration, normal RBCs, number of RBCs with hemoglobin concentrations > 28 g / dL and < 41 g / dL, number of RBCs with hemoglobin volumes > 60 fL and < 120 fL, hemoglobin distribution width, RBC mean corpuscular volume, RBC volume distribution width, total RBC count, hemoglobin content distribution width and mean hemoglobin content, and combinations thereof.

[0049] The WBC parameters include WBC types such as lymphocytes, neutrophils, basophils, monocytes, eosinophils, polymorphonuclear cells, large unstained cells (LUC), blasts, and platelets. In addition, WBC maturity parameters such as WBC count, WBC percentage, neutrophil to lymphocyte ratio, large unstained cell to lymphocyte ratio, number of WBCs showing pseudobasophilia, and delta neutrophil index are also considered. The parameters associated with the subject indicate inflammation and immune activation. Thus, the parameters may be used to determine the presence of infection in the subject. In an embodiment, the WBC parameters are preselected. In embodiments, the WBC parameters include one or more of WBC types (such as lymphocytes, neutrophils, basophils, monocytes, eosinophils, polymorphonuclear cells, large unstained cells (LUC), blasts, and platelets), WBC counts, WBC percentages, neutrophil to lymphocyte ratio, large unstained cell to lymphocyte ratio, number of WBCs showing pseudobasophilia, WBC maturity parameters such as delta neutrophil index, and combinations thereof.

[0050] In embodiments, platelet parameters are included and taken into consideration, including, but not limited to, those indicated for platelet creatinine, platelet volume, platelet count, and platelet percentage.

[0051] In step 202, thresholds or thresholds for sensitivity and specificity associated with the trained machine learning model are determined for a desired use case. In an embodiment, the thresholds are pre-determined. In an embodiment, the trained machine learning model is an ensemble of decision tree-based methods. For example, an Extra Tree Classifier and a Light Gradient Boosting Machine (LightGBM) may be used as the machine learning model. The ensemble of decision tree-based models is trained using hematology data associated with multiple subjects. The hematology data includes WBC data and RBC data obtained from a complete blood count / percentage test associated with multiple subjects. The machine learning model is trained with 5-fold cross-validation and optimized for sensitivity. The Extra Tree Classifier is an ensemble learning technique that improves accuracy and controls overfitting by learning randomized decision trees on different data subsamples. The Light Gradient Boosting Machine (LightGBM) is a boosting approach for decision tree learning in which continuous feature values ​​are discretized into bins for faster training and reduced memory usage. The different classifiers with associated hyperparameters used in the ensemble are specified below (a-g). The weights for each classifier in the ensemble in the same order as listed are [2, 1, 1, 2, 1, 1, 1]. a.ExtraTreesClassifier(bootstrap=False,class_weight=“balanced”,criterion=“gini”,max_features=“log2”,min_ samples_leaf=0.47421052631578947,min_samples_split=0.15052631578947367,n_estimators=600,oob_score=False) b.LightGBMClassifier(boosting_type=“gbdt”,colsample_bytree=0.6933333333333332,learning_rate=0.05789894736842106,subsample_for_bin=190,max_depth=3,min_child_weight=9,min_data_in_leaf=0.03793724137931035,min_split_gain=0.15789473684210525,n_estimators=50,num_leaves=230,reg_alpha=0.2631578947368421,reg_lambda=0.7894736842105263,subsample=0.3963157894736842) c.ExtraTreesClassifier(bootstrap=False,class_weight=“balanced”,criterion=“gini”,max_features=0.2,min_samples_leaf=0.01,min_samples_split=0.01,n_estimators=200,oob_score=False) d.ExtraTreesClassifier(bootstrap=True,class_weight=“balanced”,criterion=“gini”,max_features=0.4,min_samples_leaf=0.01,min_samples_split=0.15052631578947367,n_estimators=10,oob_score=True) e.ExtraTreesClassifier(bootstrap=False,class_weight=“balanced”,criterion=“gini”,max_features=“log2”,min_samples_leaf=0.01,min_samples_split=0.056842105263157895,n_estimators=400,oob_score=False) f.ExtraTreesClassifier(bootstrap=False,class_weight=“balanced”,criterion=“gini”,max_features=0.8,min_samples_leaf=0.01,min_samples_split=0.15052631578947367,n_estimators=25,oob_score=False) g.ExtraTreesClassifier(Default parameters). (As used herein, the term "Algorithm A" refers to the above algorithm).

[0052] In one embodiment, the thresholds for sensitivity and / or specificity associated with the trained machine learning model are determined based on a requirement scenario associated with a healthcare provider. For example, in an embodiment, the requirement scenario associated with a healthcare provider includes at least one of reducing RT-PCR testing burden, reducing infectious disease testing burden, replacing RT-PCR testing, and / or determining the need for testing of a subject. In the requirement scenario of reducing RT-PCR testing burden, the algorithm can be used with a moderately high sensitivity (0.80-0.95) such that most positives are predicted as positive and some negatives are predicted as positive. By selecting a desired probability threshold, RT-PCR testing can be avoided in samples predicted as negative by the algorithm. This approach can significantly reduce the RT-PCR testing burden in many laboratories, but miss a small number of positives. This use case can be useful in a pandemic scenario. Similarly, in the requirement scenario of replacing RT-PCR testing, the algorithm can be used with a very high sensitivity (0.99-1.0) such that all positives are predicted as positive and many negatives are predicted as positive. In this high sensitivity scenario, RT-PCR testing can be avoided in samples predicted as negative without missing any true positives. This approach can reduce the burden of RT-PCR testing in many laboratories / hospitals without missing any positives. This use case can be useful in pandemic scenarios. For requirement scenarios to determine the need for testing of subjects, the algorithm can be used with very high specificity (0.99-1.0) such that all negatives are predicted as negative but some positives are flagged as positives. This approach can predict positives when no one is suspected while positive flagging for negatives is minimized or nonexistent. This use case can be useful in endemic scenarios in emergency rooms.

[0053] In step 203, the trained machine learning model is used to predict the presence of the disease in the subject. Such prediction is made based on the sensitivity and specificity of the disease as well as multiple parameters associated with the subject. The binary classification model generates a probability that the input is classified into a category (in this case, COVID positive or negative). The final positive / negative callout is based on a selected threshold. The default threshold in such a model may be 0.5 (i.e., if the model outputs the probability of predicting positive for any input as ≧0.5, the input is labeled as positive, and if the probability of predicting positive is <0.5, the input is labeled as negative). Adjusting the threshold to different values ​​depending on the scenario allows for curating the output from the model. For example, if a higher positive predictive rate (fewer false positives) is required, a higher threshold for sensitivity and specificity is selected. However, if a higher negative predictive rate (fewer false negatives) is required, a lower threshold for sensitivity and specificity is selected. In step 204, the prediction is output to the output unit 105.

[0054] For example, the prediction may be displayed on a display unit. In an alternative embodiment, the method includes using a trained machine learning model to determine the need for hospitalization of a subject based on a plurality of parameters associated with the subject. The model is trained using parameters associated with a plurality of subjects who have tested positive for a disease, and the parameters associated with the plurality of subjects are obtained within 2-3 days of the subject testing positive for the disease. Further, the model is also trained with health data associated with the subject that may be observed and collected over a period of 7-10 days after the subject tests positive for the disease.

[0055] In yet another embodiment, the method can include predicting the onset of COVID-19 sequelae in a subject. The model is trained using parameters associated with a number of subjects who have tested positive for the disease, as well as health data associated with a number of subjects that may be observed and collected over the next several months after the subjects test positive for the disease.

[0056] FIG. 3 illustrates a receiver operating characteristic curve, or ROC curve 300, for selecting thresholds for sensitivity and specificity associated with a trained machine learning model for a requirements scenario, according to one embodiment. A ROC curve is a graphical plot that shows the diagnostic ability of a binary classification model as its discrimination threshold is varied. For various thresholds, the ROC plots the true positive rate (sensitivity) against the false positive rate (probability of false alarm / 1-specificity). The diagonal divides the ROC space. Points above the diagonal represent good classification results (better than random) and points below the line represent poor results (worse than random). treatment

[0057] In embodiments, the present disclosure includes a method of treating a subject in need thereof by determining that the subject is a subject in need thereof (e.g., having a pathogenic infection, or sepsis, etc.) and then treating the subject. For example, the methods of the present disclosure can be applied to diagnose a patient as positive for a pathogenic infection, sepsis, or viral infection such as COVID-19, and then administer a therapeutic agent or compound in a therapeutically effective amount to a subject in need thereof. For example, a physician can administer a therapeutically effective amount of any drug, therapeutic agent, or biologic suitable for treating a pathogenic infection or disease state after diagnosis according to the present disclosure. Non-limiting examples of drugs suitable for treating viral infections such as COVID-19 include VEKLURY® brand Remdesivir (100 MG for injectable antiviral to treat patients hospitalized with COVID-19). See, for example, International Patent Application WO2017 / 049060A1, and US Patent Application US20210395345, entitled "Methods for treating or preventing SARS-CoV-2 infections and covid-19 with anti-SARS-CoV-2 spike glycoprotein antibodies." The therapeutically effective amount of the agent provided may be determined by a physician based on the subject's response, comorbidities, and the like. In embodiments, a therapeutic compound is administered in an effective amount to a subject in a disease state. In embodiments, a therapeutically effective amount of a compound is administered to a subject in need thereof in an amount sufficient to alter the natural or presented state of the subject.

[0058] In an embodiment, the subject may be in a disease state resulting from infection with an RNA virus. Non-limiting examples of RNA viruses include segmented single-stranded negative-sense RNA viruses (e.g., orthomyxoviruses), non-segmented single-stranded negative-sense RNA viruses (mononegaviruses), non-segmented single-stranded positive-sense RNA viruses (e.g., coronaviruses), ambisense RNA viruses (e.g., bunyaviruses and arenaviruses, e.g., Lassa virus, Junin virus, Machupo virus, and lymphocytic choriomeningitis virus), and double-stranded RNA viruses (e.g., reoviruses). Other viruses that cause disease include segmented single-stranded negative-sense RNA viruses, non-segmented single-stranded negative-sense RNA viruses, non-segmented single-stranded positive-sense RNA viruses, or double-stranded RNA viruses (e.g., reoviruses). Additional RNA viruses include picornaviruses, togaviruses (e.g., sindbis virus), flaviviruses, and coronaviruses. The virus may be any type, species, and / or strain of picornavirus, togavirus, flavivirus, and coronavirus. Additional disease-causing viruses include respiratory syncytial virus (RSV), influenza virus (influenza A, B, or C), human metapneumovirus (HMPV), rhinovirus, parainfluenza virus, SARS coronavirus, human immunodeficiency virus (HIV), hepatitis virus (A, B, C), Ebola virus, herpes virus, rubella, variola major, and variola minor.

[0059] In certain embodiments, the therapeutic composition is administered to a subject diagnosed with a disease caused by infection with a virus, for example, the patient is infected with respiratory syncytial virus (RSV), influenza virus (influenza A virus, influenza B virus, or influenza C virus), human metapneumovirus (HMPV), rhinovirus, parainfluenza virus, SARS coronavirus, human immunodeficiency virus (HIV), hepatitis virus (A, B, C), Ebola virus, herpes virus, rubella, variola major, and / or variola minor.

[0060] In certain embodiments, the disease treated according to the methods described herein is a disease caused by bacterial infection. Non-limiting examples of disease-causing bacteria include Streptococcus pneumoniae, Mycobacterium tuberculosis, Chlamydia pneumoniae, Bordetella pertussis, Mycoplasma pneumoniae, Haemophilus influenzae, Mycobacterium catarrhalis, Legionella pneumoniae, Pneumocystis jiroveci, Chlamydia psittaci, Chlamydia trachomatis, Bacillus anthracis, Francisella tularensis, Borrelia burgdorferi, Salmonella enterica, Yersinia pestis, Shigella, Escherichia coli, Corynebacterium diphtheriae, and Treponema pallidum.

[0061] In certain embodiments, the composition is administered to a patient diagnosed with a disease caused by infection with bacteria, for example, the patient is infected with Streptococcus pneumoniae, Mycobacterium tuberculosis, Chlamydia pneumoniae, Bordetella pertussis, Mycoplasma pneumoniae, Haemophilus influenzae, Bordetella catarrhalis, Legionella pneumoniae, Pneumocystis jiroveci, Chlamydia psittaci, Chlamydia trachomatis, Bacillus anthracis, Francisella tularensis, Borrelia burgdorferi, Salmonella enterica, Yersinia pestis, Shigella, Escherichia coli, Corynebacterium diphtheriae, and / or Treponema pallidum.

[0062] In certain embodiments, the composition is administered to a patient diagnosed with a disease caused by a fungal infection, for example, the patient is infected with Blastomyces, Paracoccidioides, Sporothrix, Cryptococcus, Candida, Aspergillus, Histoplasma, Cryptococcus, Bipolaris, Cladophialophora, Cladosporium, Rice eyespot, Exophiala, Fonseca, Phialophora, Cladophialophora, Ochroconis, Rhinocladiella, Scolecobasidium, and / or Exophiala.

[0063] In certain embodiments, the composition is administered to a patient diagnosed with a disease caused by an infection with a yeast, for example, the patient may be infected with Acycloconidium, Botryoacus, Brettanomyces, Brera, Breromyces, Candida, Citeromyces, Clavispora, Cryptococcus, Cystophilobasidium, Debaromyces, Debaryomyces, Dekkera, Dipodacus, Endomyces, Endomycopsis, Erythrobasidium, Phellomyces, Phyllobasidium, Gilliermondella, Hanseniaspora, Hansenula, Hasegawaea, Hilopichia, Isatienchia, Chloeckella, Kluyveromyces, Komagatella, Leucosporidium, Lipomyces, Rhoderomyces, Malassezia-mastigomices, Mestiniko Infected with Bia, Murrachia, Nadsonia, Octosporomyces, Osporidium, Pachysolene, Petasospora, Phaffia, Pichia, Pseudozyma, Rhodium, Rhodotrudeura, Saccharomyces, Saccharomycodes, Saccharomycopsis, Schizosaccharomyces, Swanniomyces, Selenotyla, Sirobasidium, Sporidiobolus, Sporobolomyces, Stephanoascus, Sterigmatomyces, Syringospora, Torulaspora, Torulopsis, Tremeloides, Trichosporon, Trigonopsis, Udeniomyces, Warthomyces, Wickelhamia, Viropsis, Wingea, Yarrowia, Zygophabospora, Zygolipomyces, and / or Zygosaccharomyces.

[0064] In certain embodiments, the composition is administered to a subject infected with a parasite, e.g., the patient is infected with Babesia, Cryptosporidium, Entamoeba histolytica, Leishmania, Giardia lamblia, Plasmodium, Toxoplasma, Trichomonas, Trypanosoma, Ascaris, Cestoda, Hookworm, Brugia, Fasciola, Trichinella, Schistosoma, Taenia, Cimex lectularius, Pediculus, and / or Sarcoptes. See additional information in U.S. Patent Publication No. 20170321192 (incorporated herein by reference in its entirety).

[0065] Exemplary Numbered Embodiments Embodiment 1. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) including patient data; and a disease prediction module (110) configured to receive a plurality of parameters associated with the subject, determine a threshold for sensitivity and / or specificity associated with a trained machine learning model, and use the trained machine learning model to predict the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject, and output the prediction to an output unit (105).

[0066] Embodiment 2. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) including patient data; and a disease prediction module (110) configured to receive a plurality of parameters associated with the subject, determine a threshold for a sensitivity and / or specificity associated with a trained machine learning model, use the trained machine learning model to predict the presence of disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject, and output the prediction to an output unit (105), wherein the threshold or thresholds for the sensitivity and / or specificity associated with the trained machine learning model are determined based on a requirement scenario associated with a healthcare provider. The system (100).

[0067] Embodiment 3. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to receive a plurality of parameters associated with the subject, determine a threshold for sensitivity and / or specificity associated with a trained machine learning model, use the trained machine learning model to predict the presence of disease in the subject based on a particular threshold or thresholds of the model and the plurality of parameters associated with the subject, and output the prediction to an output unit (105), wherein the threshold or threshold for sensitivity and / or specificity associated with the trained machine learning model is determined based on a requirement scenario associated with a healthcare provider, the requirement scenario associated with the healthcare provider comprising at least one of reducing the burden of RT-PCR testing, substituting RT-PCR testing, and / or determining the need for testing of the subject. The system (100).

[0068] Embodiment 4. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to receive a plurality of parameters associated with the subject, determine thresholds for sensitivity and / or specificity associated with a trained machine learning model, use the trained machine learning model to predict the presence of a disease in the subject based on certain thresholds of the model and the plurality of parameters associated with the subject, and output the prediction to an output unit (105), wherein the plurality of parameters associated with the subject include red blood cell (RBC) parameters and white blood cell (WBC) parameters.

[0069] Embodiment 5. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) including patient data; a plurality of parameters associated with the subject; determining thresholds for sensitivity and / or specificity associated with a trained machine learning model; predicting the presence of the disease in the subject based on certain thresholds of the model and the plurality of parameters associated with the subject, and outputting the prediction to an output unit (105). and a configured disease prediction module (110), wherein the plurality of parameters associated with the subject include red blood cell (RBC) parameters and white blood cell (WBC) parameters, and the RBC parameters include one or more of hemoglobin level, hematocrit, RBC size, hemoglobin level in the individual's RBCs, mean corpuscular volume, mean corpuscular hemoglobin concentration, normal RBCs, number of RBCs having hemoglobin concentrations > 28 g / dL and < 41 g / dL, number of RBCs having hemoglobin volumes > 60 fL and < 120 fL, hemoglobin distribution width, RBC mean corpuscular volume, RBC volume distribution width, total RBC count, hemoglobin content distribution width, and average hemoglobin content.

[0070]

[0023] Embodiment 6. A system (100) for predicting the presence of a disease in a subject, the system (100) receiving a plurality of parameters associated with the subject, the plurality of parameters including one or more processing units (101) and a medical database (112) coupled to the one or more processing units (101), the medical database (112) including patient data, determining thresholds for sensitivity and / or specificity associated with a trained machine learning model, and using the trained machine learning model to predict the presence of a disease in a subject based on a particular threshold of the model and a particular threshold of the subject. and a disease prediction module (110) configured to predict the presence of disease in a subject based on a plurality of parameters associated with the body and output the prediction to an output unit (105), wherein the plurality of parameters associated with the subject include red blood cell (RBC) parameters and white blood cell (WBC) parameters, and the WBC parameters include one or more of WBC type, WBC count, WBC percentage, neutrophil to lymphocyte ratio, large unstained cell to lymphocyte ratio, number of WBCs showing pseudobasophilia, and WBC maturity parameters.

[0071] Embodiment 7. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) including patient data; and a disease prediction module (110) configured to receive a plurality of parameters associated with the subject, determine thresholds for sensitivity and / or specificity associated with a trained machine learning model, use the trained machine learning model to predict the presence of the disease in the subject based on certain thresholds of the model and the plurality of parameters associated with the subject, and output the prediction to an output unit (105), the plurality of parameters associated with the subject being determined. 20. A system (100) comprising: a blood glucose measuring system (100) for measuring a blood glucose level in an individual; a blood glucose level measuring a blood glucose level in an individual; a blood glucose level measuring a blood glucose level in an individual; a blood glucose level measuring a blood glucose level in an individual;

[0072] Embodiment 8. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) including patient data; and a disease prediction module (110) configured to receive a plurality of parameters associated with the subject, determine thresholds for sensitivity and / or specificity associated with a trained machine learning model, use the trained machine learning model to predict the presence of the disease in the subject based on certain thresholds of the model and the plurality of parameters associated with the subject, and output the prediction to an output unit (105), wherein the trained machine learning model is an ensemble of decision tree-based methods.

[0073] Embodiment 9. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) including patient data; and a disease prediction module (110) configured to receive a plurality of parameters associated with the subject, determine thresholds for sensitivity and / or specificity associated with a trained machine learning model, use the trained machine learning model to predict the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject, and output the prediction to an output unit (105), wherein the disease prediction module (110) is further configured to use the trained machine learning model to predict the need for hospitalization of the subject based on the plurality of parameters associated with the subject.

[0074] Embodiment 10. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to receive a plurality of parameters associated with the subject, determine thresholds for sensitivity and / or specificity associated with a trained machine learning model, and use the trained machine learning model to predict the presence of disease in the subject based on certain thresholds of the model and the plurality of parameters associated with the subject, and output the prediction to an output unit (105), wherein the disease prediction module (110) is further configured to predict the onset of COVID-19 sequelae in the subject.

[0075] Embodiment 11. A method (200) for predicting the presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with a trained machine learning model; predicting the presence of the disease in the subject using the trained machine learning model based on a particular threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction to an output unit.

[0076] Embodiment 12. A method (200) for predicting the presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with a trained machine learning model; predicting the presence of the disease in the subject using the trained machine learning model based on a particular threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction to an output unit, wherein the threshold for sensitivity and / or specificity associated with the trained machine learning model is determined based on a requirement scenario associated with a healthcare provider.

[0077] Embodiment 13. A method (200) for predicting the presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with a trained machine learning model; predicting the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject using the trained machine learning model; and outputting the prediction to an output unit, wherein the threshold for sensitivity and / or specificity associated with the trained machine learning model is determined based on a requirement scenario associated with a healthcare provider, the requirement scenario associated with the healthcare provider comprising at least one of reducing the burden of RT-PCR testing, substituting RT-PCR testing, and / or determining the need for testing of the subject.

[0078] Embodiment 14. A method (200) for predicting the presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with a trained machine learning model; predicting the presence of the disease in the subject using the trained machine learning model based on a particular threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction to an output unit, wherein the disease is COVID-19.

[0079] Embodiment 15. A method (200) for predicting the presence of disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with a trained machine learning model; predicting the presence of disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject using the trained machine learning model; and outputting the prediction to an output unit, the method (200) further comprising: predicting the need for hospitalization of the subject based on the plurality of parameters associated with the subject using the trained machine learning model; and predicting the onset of COVID-19 sequelae in the subject.

[0080] Embodiment 16. A method (200) for predicting the presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with a trained machine learning model; predicting the presence of the disease in the subject using the trained machine learning model based on a particular threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction to an output unit, wherein the disease is COVID-19.

[0081] Embodiment 17. A method (200) for predicting the presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with a trained machine learning model; predicting the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject using the trained machine learning model; and outputting the prediction to an output unit, the method (200) further comprising: predicting the need for hospitalization of the subject based on the plurality of parameters associated with the subject using the trained machine learning model; and predicting the onset of a viral, bacterial, or fungal infection in the subject.

[0082] Embodiment 18. An article of manufacture, such as a system or a component thereof, including a non-transitory computer-readable medium having instructions encoded thereon, the instructions being configured to cause one or more processors to perform the methods of embodiments 15-17 above.

[0083] Embodiment 19. In an embodiment, the present disclosure includes a non-transitory computer readable medium having instructions encoded thereon, the instructions configured to cause one or more processors to execute a method including predicting the presence of a disease in a subject, the method (200) including receiving a plurality of parameters associated with the subject, determining a threshold for sensitivity and / or specificity associated with a trained machine learning model, using the trained machine learning model to predict the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject, and outputting the prediction to an output unit, further including using the trained machine learning model to predict the need for hospitalization of the subject based on the plurality of parameters associated with the subject, and predicting the onset of a viral, bacterial, or fungal infection in the subject.

[0084] Embodiment 20. In an embodiment, the present disclosure includes a non-transitory computer readable medium having instructions encoded thereon, the instructions configured to cause one or more processors to execute a method including predicting the presence of a disease in a subject, the method (200) including receiving a plurality of parameters associated with the subject, determining a threshold for sensitivity and / or specificity associated with a trained machine learning model, using the trained machine learning model to predict the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject, and outputting the prediction to an output unit, and further including using the trained machine learning model to predict the need for hospitalization of the subject based on the plurality of parameters associated with the subject, and predicting the onset of sepsis in the subject.

[0085] Embodiment 21. An article of manufacture including a non-transitory computer-readable medium having instructions encoded thereon, the instructions configured to cause one or more processors to execute a method including predicting the presence of a disease according to an embodiment of the present disclosure.

[0086] Embodiment 22. A non-transitory computer-readable medium having instructions encoded thereon, the instructions being configured to cause one or more processors to execute a method including predicting the presence of a disease according to at least one embodiment of the present disclosure.

[0087] Embodiment 23. A method of treating a subject in need thereof, comprising predicting the presence of a disease in a subject according to an embodiment of the present disclosure and subsequently treating the subject in need thereof. In an embodiment, treating comprises administering a therapeutically effective amount of a compound to the subject.

[0088] Embodiment 24. A method of treating a subject in need thereof, including predicting the presence of a disease in the subject, the method (200) comprising receiving a plurality of parameters associated with the subject, determining thresholds for sensitivity and / or specificity associated with a trained machine learning model, predicting the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject using the trained machine learning model, outputting the prediction to an output unit, and thereafter treating the subject in need thereof. In an embodiment, treating comprises administering a therapeutically effective amount of a compound to the subject. In an embodiment, the disease is characterized as a viral disease, a bacterial disease, COVID-19, COVID sequelae, or the like.

[0089] Embodiment 25. In an embodiment, the present disclosure includes a system for predicting the presence of a disease in a population. In one aspect of the present disclosure, the system includes a processing unit and a memory. Further, the memory includes a disease prediction module configured to receive a plurality of parameters associated with one or more subjects and determine a threshold value for sensitivity and / or specificity associated with the trained machine learning model. Further, the module is configured to predict the presence of a disease in one or more subjects or a population of subjects and the plurality of parameters associated with the subject or population of subjects using the trained machine learning model, and output the prediction to an output unit. EXAMPLES

[0090] Example 1 This example describes a COVID-19 detection algorithm based on CBC / Diff hematology data. In an embodiment, the disclosure includes, among other things, a multi-parametric machine learning algorithm to run on a hematology instrument and provide a COVID-19 flag along with CBC / Diff data. In an embodiment, the algorithm can be run with different sensitivity and specificity values ​​based on the intended use case, in a pandemic or endemic setting. The training data used in the algorithm is based on an unvaccinated cohort from 2020. Additional data from a greater variety of positive and negative individuals may be used to improve the robustness of the algorithm.

[0091] Blood tests such as CBC / Diff (Complete Blood Count with Differential) are ubiquitously performed for overall health monitoring. In an embodiment, the present disclosure combines CBC / Diff results from a Siemens Healthineers ADVIA® 2120i Hematology System with a multi-parametric machine learning algorithm for rapid flagging of COVID-19 (Coronavirus Disease 2019) positives. The generation of diagnostic data and the execution of the algorithm on the data all occur on the same instrument, ensuring easy data access. Additionally, the algorithm can also run on a connected laboratory information system. Such an algorithm can be beneficial when access to COVID testing is limited and for predicting COVID-19 in unsuspected individuals to prompt confirmatory testing. Various pandemic and endemic applications for this algorithm are demonstrated here. Methods: CBC / Diff data for 309 COVID-19 positives and 245 negatives were obtained from a 2020 pre-vaccination cohort from Mumbai, India. Using this data, a machine learning classification model was trained with 5-fold cross-validation to distinguish COVID-19 positives from negatives. Results: The final ML model was an ensemble of decision tree-based methods with an accuracy of 74% and an AUROC of 0.79 on the test set. The most significant features included eosinophil and lymphocyte counts as previously reported. The model was able to more accurately predict hospitalized COVID-19 patients (92% correctly classified) than non-hospitalized patients (70% correctly classified). Conclusions A predictive algorithm, method, and system for rapid COVID-19 flagging using data from routine CBC testing is provided. The algorithm may be useful in pandemic or endemic scenarios as an alternative to testing, reducing testing burden and flagging non-suspicious positives.

[0092] Consideration Hematology data provides a snapshot of the patient's health status and represents the host's response to infection. Moreover, blood tests such as CBC / Diff are commonly prescribed before suspicion of any disease and are therefore an ideal platform to flag and stratify possible infections or health conditions, thus allowing earlier intervention and streamlining different medical resources (see, e.g., Barnes PW, McFadden SL, Machin SJ, Simson E. The international consensus group for hematology review: suggested criteria for action following automated CBC and WBC differential analysis. Laboratory hematology: official publication of the International Society for Laboratory Hematology 2005;11 2:83-90). Here, the CBC / Diff test performed on a Siemens Healthineers ADVIA® 2120i Hematology System was combined with a multi-parametric AI algorithm to flag COVID-19 positivity in individuals.

[0093] Studies have reported hematological features such as eosinopenia and lymphopenia associated with COVID-19 positivity and prognosis (e.g., Frater JL, Zini G, d'Onofrio G, Rogers HJ. COVID-19 and the clinical hematology laboratory. International Journal of Laboratory Hematology 2020;42:11-8, Guan Wj, Ni Zy, Hu Y, Liang Wh, Ou Cq, He Jx, Liu L et al. Clinical characteristics of coronavirus disease 2019 in China. New England journal of medicine 2020;382:1708-20, Henry BM, De Oliveira MHS, Benoit S, Plebani M, Lippi G. Hematologic, biochemical and immune biomarker abnormalities associated with severe illness and mortality in coronavirus disease 2019(COVID-19):a meta-analysis}. Clinical Chemistry and Laboratory Medicine(CCLM)2020;58:1021, Jiang SQ, Huang QF, Xie WM, Lv C, Quan XQ.The association between severe COVID-19 and low platelet count:evidence from 31 observational studies involving 7613 participants.British journal of haematology 2020:e29-e33, and Kermali M, Khalsa RK, Pillai K, Ismail Z, Harky A. The role of biomarkers in diagnosis of COVID-19--A systematic review.Life sciences 2020;254:117788). Some hematology parameters also show mixed trends in different studies, possibly due to cohort dependency of parameter values. However, it is difficult for physicians to identify COVID-19 positivity by evaluating a single parameter in isolation, or a combination of parameters. Multiparametric machine learning (ML) algorithms allow modeling of complex relationships between multiple predictive parameters to generate predictions more accurately and efficiently. Unlike multiparametric studies that combine parameters measured on multiple instruments (see, for example, Kukar M, Guncar G, Vovko T, Podnar S, Chernelck P, Brvar M, Zalaznik M et al. COVID-19 diagnosis by routine blood tests using machine learning. Scientific reports 2021;11:1-9), the present disclosure makes predictions using multiple hematology parameters measured / derived using a single instrument. This eliminates the need to aggregate data from multiple sources and allows easy data access without missing values, a common problem in such multiparametric algorithms. .

[0094] While COVID-19 testing using RT-PCR (reverse transcriptase polymerase chain reaction) and antigen testing has been scaled up massively since the onset of the pandemic, there are multiple use cases where a CBC / Diff-based predictive algorithm can be applied to the clinical value chain in both pandemic and endemic scenarios. First, it can serve as a testing alternative in countries dealing with unavailability or shortage of such tests. Second, rapid flagging by a highly sensitive hematology-based machine learning (ML) algorithm can reduce the testing burden in pandemic scenarios. Third, an algorithm with a high negative predictive value (NPV) can reduce the number of patients who are tested prior to a procedure like surgery, thereby reducing the time to patient care. Fourth, in endemic scenarios, an algorithm with high specificity can be used to flag asymptomatic individuals who can be tested to confirm infection. The same ML algorithm with different classification thresholds can be used in multiple scenarios as described above.

[0095] Hematology-based multi-parametric ML algorithms can be used for COVID flagging within seconds to minutes. The resulting time reduction can avoid unnecessary isolation of negatives and speed access to care for patients who need to undergo procedures like surgery in a pandemic scenario, and can also be used to flag unsuspected positives during endemic scenarios.

[0096] The present disclosure includes a multi-parametric machine learning algorithm for predicting COVID-19 positivity in individuals using CBC / Diff data derived from a single device. In an embodiment, the multi-parametric ML algorithm predicts COVID-19 positivity in individuals with 74% accuracy and an AUROC (area under the receiver operating characteristic curve) of 0.79. Furthermore, in an embodiment, the algorithm predicts positive hospitalized patients with 91.7% accuracy, lending credence to our algorithm performance and the potential of the selected parameters.

[0097] material and method Dataset CBC / Diff data were obtained from two ADVIA® 2120i instruments installed at Dr. Jariwala Laboratory & Diagnostics, Mumbai (India) for this analysis. This disclosure involves a retrospective study of a cohort of 309 positive and 245 COVID-19 negative individuals, determined using RT-PCR testing. Only anonymized data was transmitted to the researchers, which included information on the gender and age of the individuals as well as disease outcomes in COVID-19 positive individuals (whether they were hospitalized / admitted to ICU / survived or died). COVID-19 positivity and negativity were measured simultaneously during the first wave of infection in India (May 2020-January 2021) before any vaccination drive activity. COVID-19 negatives included in this study were individuals who received medical attention for other illnesses during the first wave of the pandemic. Of the 309 positives, 60 were hospitalized at the time of CBC / Diff testing, of which 29 were admitted to ICU, and 53 survived the infection. The demographics and blood counts of COVID-19 positive and negative individuals included in this study are summarized in Table 1. [Table 1]

[0098] Parameter Selection The starting dataset included all parameters included in the ADVIA 2120i CBC / Diff raw dataset. Parameters were excluded using various criteria, e.g., if the median did not change significantly between positive and negative (performed using box plots and p-values ​​for Kruskal-Wallis H-test), parameters that measured only device characteristics, etc. Redundant and highly correlated parameters were excluded (Spearman correlation coefficient >0.9). Finally, a set of 48 parameters was identified that included both raw and calculated patient parameters (see Figure 5). All 48 parameters were numeric and no missing values ​​were present. No patient demographic information or information identifying the source device (Device 1 or 2) was included while training the algorithm.

[0099] algorithm After parameter identification, Microsoft Azure AutoML (https: / / www.microsoft.com / en-us / research / project / automl / ) was used to train a classification model to distinguish between positives and negatives. 80% of the data (247 positives, 196 negatives) were used to train the model using 5-fold cross-validation, and the remaining 20% ​​of the data (62 positives, 49 negatives) was used as a test set. Azure AutoML uses the dataset to automatically train many different models, and model training can be optimized based on different metrics. When optimized for sensitivity, a good model for the dataset is a voting ensemble of multiple tree-based classifiers. The different classifiers with associated hyperparameters used in the ensemble are specified below (a-g). The weights per classifier in the ensemble in the same order as listed are [2, 1, 1, 2, 1, 1, 1]. h.ExtraTreesClassifier(bootstrap=False,class_weight=“balanced”,criterion=“gini”,max_features=“log2”,min_samples_leaf=0.47421052631578947,min_samples_split=0.15052631578947367,n_estimators=600,oob_score=False) i.LightGBM Classifier(boosting_type=“gbdt”,colsample_bytree=0.6933333333333332,learning_rate=0.05789894736842106,subsample_for_bin=190,max_depth=3,min_child_weight=9,min_data_in_leaf=0.03793724137931035,min_split_gain=0.15789473684210525,n_estimators=50,num_leaves=230,reg_alpha=0.2631578947368421,reg_lambda=0.7894736842105263,subsample=0.3963157894736842) j.ExtraTreesClassifier(bootstrap=False,class_weight=“balanced”,criterion=“gini”,max_features=0.2,min_samples_leaf=0.01,min_samples_split=0.01,n_estimators=200,oob_score=False) k.ExtraTreesClassifier(bootstrap=True,class_weight=“balanced”,criterion=“gini”,max_features=0.4,min_samples_leaf=0.01,min_samples_split=0.15052631578947367,n_estimators=10,oob_score=True) l.ExtraTreesClassifier(bootstrap=False,class_weight=“balanced”,criterion=“gini”,max_features=“log2”,min_samples_leaf=0.01,min_samples_split=0.056842105263157895,n_estimators=400,oob_score=False) m.ExtraTreesClassifier(bootstrap=False,class_weight=“balanced”,criterion=“gini”,max_features=0.8,min_samples_leaf=0.01,min_samples_split=0.15052631578947367,n_estimators=25,oob_score=False) n.ExtraTreesClassifier(Default parameters).

[0100] The trained ensemble models were downloaded from Microsoft Azure, which allows great flexibility in the use of the models, e.g., evaluating model performance on a test set and deploying them.

[0101] result The performance of the models was evaluated based on various metrics including predictive accuracy, AUROC, sensitivity, and specificity. Then, feature importance obtained using the MIMIC LightGBM explainer on the training data was presented. The MIMIC LightGBM explainer approximates a complex black-box ensemble model obtained from Azure AutoML using the LightGBM surrogate model to obtain feature / parameter importance values. A comparison was made between the results from our 48-parameter model and other models trained using the most important parameters.

[0102] Different use cases for the algorithm Model evaluation The entire data set was split into a separate training set (80% of the total data) and a test set (the remaining 20% ​​of the data). The model was trained using 5-fold cross-validation on the training set, and model performance was evaluated on the test set. The test dataset consisted of 62 positives and 49 negatives. The model performed with an accuracy of 74%, a sensitivity of 0.73, a specificity of 0.74, and an AUROC of 0.79 on the test data (all metrics reported using a default classification threshold of 0.5). Figure 4 shows the ROC (Receiver Operating Curve) curve calculated using the test data. Since the positives and negatives in the test are balanced, the AUROC is a good performance metric for validation.

[0103] Of the 62 positives in the test set, 12 were hospitalized, whereas the remaining 50 were not. We observed that hospitalized positivity was predicted more accurately than non-hospitalized positivity (Table 2). [Table 2]

[0104] This increases the confidence in the selected parameters.,The effects of age and gender were also observed on the,algorithm performance (see Supplementary Table 1 and Supplementary,Table 2 below), and it was found that the algorithm,predictions did not show any bias in age or gender.

[0105] Feature Importance Here, the model explanations obtained using the MIMIC LightGBM explainer on the training data are presented. Figure 5 shows the features in order of increasing importance value calculated by the model explainer.

[0106] Comparison with models generated with higher parameters The importance of the selected 48 parameters was demonstrated by training two other classification models using the top 20 and top 10 most important features obtained from the MIMIC explainer (Figure 5). The performance of the three models on the test set is summarized in Table 3. [Table 3]

[0107] The performance of the top 20 parameter models was observed to be inferior to the 48 parameter model in terms of both overall accuracy and sensitivity (or % correctly predicted positives). The predictive accuracy drops even further for the top 10 parameter models, especially for specificity (or % correctly predicted negatives). These results indicate that all 48 parameters are indeed important in distinguishing positives from negatives.

[0108] Use Cases A hematology-based ML algorithm for COVID-19 flagging can be useful in multiple use cases in pandemic and endemic scenarios. In embodiments, the classification threshold can be adjusted to utilize the same predictive model useful in different settings. The present disclosure shows two use cases in a pandemic scenario and one more in an endemic scenario (Figures 6A-6C).

[0109] In countries facing inaccessibility or shortage of COVID tests in a pandemic scenario, AI-based COVID-19 flagging using CBC / Diff data can serve as a low-cost and rapid alternative. As shown in Figure 6A, the algorithm can be used with a sensitivity of 0.94 and a specificity of 0.39 (point 1 in the ROC curve) to make predictions for this use case. Depending on the selected sensitivity, the algorithm can capture the majority of positives. This result can be used as an additional point in clinical decision-making in the absence of confirmatory testing.

[0110] In another application in a pandemic scenario, the algorithm can be used with high sensitivity to reduce the burden of infection testing. This can be especially useful in large laboratories. The rapid turnaround time of CBC / Diff results on the ADVIA 2120i analyzer (~1 min) along with the algorithm execution time (seconds) provides a faster alternative to RT-PCR or antigen testing. At point 2 (Figure 6A) in the ROC curve, the algorithm has an NPV of 1, where negatives are predicted with high certainty, no testing is required, and quarantine can be exempted. Individuals predicted as positive by the algorithm need to be followed up with COVID-specific testing. This approach can reduce the testing burden by about 10%.

[0111] The algorithm can also be used in endemic scenarios to drive testing in individuals with fever in the emergency department (ED) when COVID-19 is not suspected. An algorithm that operates with high specificity and a positive predictive value (PPV) of 1 (point 3 on the ROC curve) can predict with high certainty COVID-19 positivity in individuals who would otherwise be missed and recommend them for COVID testing. Importantly, with a PPV of 1, negatives are not flagged for testing, eliminating false alarms.

[0112] Consideration An embodiment provides a multi-parametric machine learning algorithm for flagging COVID-19 positivity in an individual using parameters obtained from a CBC / Diff test performed on a Siemens Healthineers ADVIA® 2120i Hematology System. The CBC / Diff test is a routinely performed blood test that can aid in the detection of various health conditions. Combining the results of the CBC / Diff test with an AI algorithm for flagging infectious diseases has several advantages. First, all parameters are obtained on a single instrument, thereby eliminating the need to aggregate data from multiple sources. Second, such an algorithm can serve as a faster alternative to standard virus-specific tests, as results can be obtained within minutes. Third, machine learning algorithms can help improve the accuracy of diagnosis as a result of learning the complex relationships between multiple parameters.

[0113] There are several studies reporting altered hematological parameters in the presence of COVID-19 infection (e.g., Henry BM, De Oliveira MHS, Benoit S, Plebani M, Lippi G. Hematologic, biochemical and immune biomarker abnormalities associated with severe illness and mortality in coronavirus disease 2019 (COVID-19), a meta-analysis). Clinical Chemistry and Laboratory Medicine (CCLM), 2020;58:1021-8, Kermali M, Khalsa RK, Pillai K, Ismail Z, Harky A. The role of biomarkers in diagnosis of COVID-19-A systematic review. Life sciences 2020;254:117788, Lippi G, Plebani M. Laboratory abnormalities in patients with COVID-2019 infection. Clinical chemistry and laboratory medicine (CCLM), 2020;58:1131-4, Wang D, Hu B, Hu C, Zhu F, Liu X, Zhang J, Wang B, et al., Clinical Characteristics of 138 Hospitalized Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan, China.JAMA 2020;323:1061~9, Yip CY, Yap ES, De Mel S, Teo WZ, Lee CT, Kan S, Lee MC, et al., Temporal changes in immune blood cell parameters in COVID-19 infection and recovery from severe infection.British Journal of Hematology 2020:33~6, Zhang Jj, Dong X, Cao Yy, Yuan Yd, Yang Yb, Yan Yq, Akdis CA, et al. Clinical characteristics of 140 patients infected with SARS-CoV-2 in Wuhan, China.Allergy 2020;75:1730~41, an d Zhang J, Wang X, Jia X, Li J, Hu (See K, Chen G, Wei J et al., Risk factors for disease severity, unimprovement, and mortality in COVID-19 patients in Wuhan, China. Clinical microbiology and infection 2020;26:767–72). Another study uses a linear combination of blood count parameters to determine a COVID-19 prognostic score (see, e.g., Linssen J, Ermens A, Berrevoets M, Seghezzi M, Previtali G, Russcher H, Verbon A, et al. A novel haemocytometric COVID-19 prognostic score developed and validated in an observational multicentre European hospital-based study. Elife;9:e63195).

[0114] In contrast, multi-parametric machine learning algorithms can learn complex non-linear relationships between multiple parameters and improve the overall performance of a diagnostic test.

[0115] An embodiment includes an ensemble of tree-based algorithms for predicting COVID-19 positive or negative status using hematology data. In an embodiment, the classifier predicts COVID-19 positivity with 74% accuracy across test sets and 91.7% accuracy in hospitalized patients. The improved performance on hospitalized patients lends greater confidence to the parameter selection process of the present disclosure. As described in Section 3.4, an embodiment of the algorithm can be advantageously used in different contexts in pandemic and endemic scenarios, including its use as a faster alternative to RT-PCR testing, reducing testing burden in pandemics, and detecting COVID-19 in unsuspected asymptomatic individuals. Another study (see Kukar M, Guncar G, Vovko T, Podnar S, Chernelck P, Brvar M, Zalaznik M, et al., COVID-19 diagnosis by routine blood tests using machine learning. Scientific reports 2021;11:1-9) uses routine blood parameters for COVID-19 diagnosis using machine learning algorithms, but those parameters need to be collected from multiple different tests, in contrast to the present embodiment, which uses parameters obtained from a single instrument. Kukar et al. also includes an asymmetric dataset (3% positive, 97% negative). Although the prevalence of COVID-19 at the time of its publication was 3%, it is believed that AUROC is not the right metric for a highly asymmetric dataset, and perhaps some other metric, such as the area under the precision-recall curve (AUPRC), may have been a better measure for reporting cross-validation results. On the other hand, the present disclosure includes a holdout test dataset (20% of the total data) with a balanced number of positives and negatives to report AUROC and other metrics.

[0116] In an embodiment, the algorithm described and used herein includes a total of 48 parameters, including percentages of various white blood cell types such as lymphocytes, neutrophils, basophils, monocytes, eosinophils, polymorphonuclear cells, large unstained cells (LUC), and blasts. In an embodiment, the parameter set also included platelet count and several red blood cell (RBC) parameters such as hemoglobin level, mean corpuscular volume (MCV), mean cellular hemoglobin concentration (g / dL), number of red blood cells with low, normal, and high hemoglobin concentration and volume, hemoglobin distribution width (HDW), which is the standard deviation of the hemoglobin concentration (HC) histogram and is expressed in g / dL, red blood cell distribution width (RDW), and total RBC count. In an embodiment, the parameter set also included several raw instrument measurements specific to the ADVIA 2120i analyzer. In addition to the parameters acquired using the ADVIA 2120i, in embodiments, two derived parameters were included in the analysis: neutrophil-to-lymphocyte ratio (NLR) and LUC-to-lymphocyte ratio. LUC is a parameter unique to the ADVIA 2120i analyzer.

[0117] Several parameters used in the algorithm of the present disclosure have been widely reported to be important in the diagnosis and / or prognosis of COVID-19. Eosinopenia has been identified in the literature as a key feature of COVID-19 patients (see, e.g., Lippi G, Plebani M. Laboratory abnormalities in patients with COVID-2019 infection. Clinical chemistry and laboratory medicine (CCLM) 2020;58:1131-4 and Zhang Jj, Dong X, Cao Yy, Yuan Yd, Yang Yb, Yan Yq, Akdis CA, et al. Clinical characteristics of 140 patients infected with SARS-CoV-2 in Wuhan, China. Allergy 2020;75:1730-41), while lymphocytes have been observed to be important biomarkers in the diagnosis and prognosis of COVID-19 (see, e.g., Henry BM, De Oliveira MHS, Benoit S, Plebani M, Lippi G. Hematologic, biochemical and immune biomarker abnormalities associated with severe illness and mortality in coronavirus disease). 2019(COVID-19):a meta-analysis}.Clinical Chemistry and Laboratory Medicine(CCLM), 2020;58:1021~8, Kermali M, Khalsa RK, Pillai K, Ismail Z, Harky A.The role of biomarkers in diagnosis of COVID-19-A systematic review.Life sciences 2020;254:117788, Lippi G, Plebani M.Laboratory abnormalities in patients with COVID-2019 infection.Clinicalchemistry and Laboratory Medicine(CCLM) 2020;58:1131~4 Wang D, Hu B, Hu C, Zhu F, Liu X, Zhang J, and Wang B Pneumonia in Wuhan, China.JAMA 2020;323:1061~9 Zhang JJ, Dong X, Cao Yy, Yuan Yd, Yang Yb, Yan Yq, Akdis CA Wuhan、China.Allergy 2020;75:1730~41、Khartabil T、Russcher H、van der Ven A、De Rijke YA 2020;57:415~31 Qin C, Zhou L, Hu Z, Zhang S, Yang S, Tao Y, and Xie C. Dysregulation of immune response in patients with coronavirus 2019 (COVID-19) in Wuhan, China.Clinical Infectious Diseases 2020;71:762~8,100.2013.2016.2014.2014.2012.0023 [ PubMed ] Qin C, Zhou L, Hu Z, Zhang S, Yang S, Tao Y, and Xie C. Dysregulation of immune response in patients with coronavirus 2019(COVID-19) in Wuhan、China.Clinical infectious diseases 2020;71:762-8) reported decreased percentages of monocytes, eosinophils, and basophils in severe cases. Leukopenia was observed in 33.7% of hospitalized patients upon admission (see, e.g., Guan Wj, Ni Zy, Hu Y, Liang Wh, Ou Cq, He Jx, Liu L, et al., Clinical characteristics of coronavirus disease 2019 in China. New England journal of medicine 2020;382:1708-20). The prevalence of thrombocytopenia has been widely reported in COVID-19 patients, especially in those with more severe disease (e.g., Henry BM, De Oliveira MHS, Benoit S, Plebani M, Lippi G. Hematologic, biochemical and immune biomarker abnormalities associated with severe illness and mortality in coronavirus disease 2019(COVID-19):a meta-analysis}. Clinical Chemistry and Laboratory Medicine(CCLM)2020;58:1021~8, Jiang SQ, Huang QF, Xie WM, Lv C, Quan XQ. The association between severe COVID-19 and low platelet count:evidence from 31 observational studies involving 7613 participants.British journal of hematology 2020:e29~e33, Kermali M, Khalsa RK, Pillai K, Ismail Z, Harky A. The role of biomarkers in diagnosis of COVID-19-A systematic review. review.Life sciences 2020;254:117788, Lippi G, Plebani M.Laboratory abnormalities in patients with COVID-2019 infection.Clinical chemistry and laboratory medicine(CCLM)2020;58:1131~4, and Lippi G, Plebani M, Henry BM.Thrombocytopenia is associated with severe coronavirus disease 2019(COVID-19)infections:a meta-analysis.Clinica chimica acta 2020;506:145~8). Several studies have also reported the impact of COVID-19 on RBC parameters such as hemoglobin, RDW, and MCV, which are also used in the algorithm of this disclosure (see, e.g., Lippi G, Plebani M. Laboratory abnormalities in patients with COVID-2019 infection. Clinical chemistry and laboratory medicine (CCLM) 2020;58:1131-4; Khartabil T, Russcher H, van der Ven A, De Rijke YA summary of the diagnostic and prognostic value of hemocytometry markers in COVID-19 patients. Critical reviews in clinical laboratory sciences 2020;57:415-31; and Yu H, Li D, Deng Z, Yang Z, Cai J, Jiang L, Wang K, et al. Total protein as a biomarker for predicting coronavirus disease-2019 pneumonia. Preprint with The Lancet 2).

[0118] This disclosure demonstrates an embodiment for detecting COVID-19 positivity using hematology parameters derived from a single device. This is a specific case study of concurrent COVID-19 positives and negatives from Mumbai, India during the first wave of infections in India. The results are excellent.

[0119] The study also highlights the use of common non-specific tests such as CBC / Diff to flag infections and alert physicians to potential health conditions of a patient. This is possible due to the unique feature of hematology instruments that measure multiple blood parameters simultaneously, the values ​​of which may change depending on the patient's health condition. In embodiments, such built-in algorithms within the hematology instrument are useful as an early warning system for physicians and aid in the management of the patient.

[0120] Supplement to Example 1 Here, the training data used for algorithm development is asymmetric with respect to gender, containing approximately twice as many males as females. The positive cohort contains 201 males and 108 females, while the negative cohort contains 116 males and 129 females. Of these, the test set consists of 43 males and 19 females in the COVID-19 positive cohort, and 22 males and 27 females in the negative cohort, respectively. However, model performance was consistent across males and females, both with a prediction accuracy of approximately 74%, thus suggesting no gender-related bias in prediction (Supplementary Table 1). [Table 4]

[0121] To understand the effect of age on algorithm performance, the predictive accuracy of embodiments of the present disclosure was compared across test sets of different age groups (Supplementary Table 2). [Table 5]

[0122] Variation in prediction accuracy was observed across different age groups. This may be due to the small and unequal number of individuals in the test set across different age groups, thus making it difficult to ascertain whether age influences algorithm performance. However, the predictive performance of the algorithm for the age group 40–80 (corresponding to the majority of the data) is fairly consistent.

[0123] The foregoing examples are provided for illustrative purposes only and should not be construed as limiting the invention disclosed herein. While the invention has been described with reference to various embodiments, it is understood that the words used herein are words of description and illustration, rather than words of limitation. Moreover, while the invention has been described herein with reference to particular means, materials, and embodiments, the invention is not intended to be limited to the details disclosed herein, but rather the invention extends to all functionally equivalent structures, methods, and uses, as fall within the scope of the appended claims. Those skilled in the art, having the benefit of the teachings herein, may effect numerous modifications thereto, and changes may be made thereto without departing from the scope and spirit of the invention within its aspects.

[0124] Example 2 Tabletop example Blood tests such as CBC / Diff (Complete Blood Count with Differential) are ubiquitously performed for overall health monitoring. In an embodiment, the present disclosure combines CBC / Diff results from a Siemens Healthineers ADVIA® 2120i Hematology System with a multi-parametric machine learning algorithm for rapid flagging of pathogen disease positives, such as positive for viral, bacterial, or fungal diseases. The generation of diagnostic data and the execution of the algorithm on the data are performed on the same instrument to ensure easy data access. Additionally, the algorithm can also be executed on a connected laboratory information system. Such algorithms can be beneficial when access to pathogen testing is limited and for predicting disease states in unsuspected individuals to prompt confirmatory testing. Various pandemic and endemic applications for this algorithm are demonstrated. Methods: Pathogen-positive and -negative CBC / Diff data were obtained from a preselected cohort. Using these data, a machine learning classification model was trained with 5-fold cross-validation to distinguish positives from negatives. Results: The final ML model is an ensemble of decision tree-based methods with good accuracy and adequate AUROC on the test set. Key features include eosinophil and lymphocyte counts as previously reported. The model is able to predict hospitalized patients more accurately than non-hospitalized patients. Conclusions: A predictive algorithm, method, and system for rapid pathogen disease flagging using data from routine CBC testing is provided. The algorithm is useful in pandemic or endemic scenarios as an alternative to testing, reducing testing burden and flagging unsuspected positives.

[0125] Example 3 Tabletop example Blood tests such as CBC / Diff (Complete Blood Count with Differential) are ubiquitously performed for overall health monitoring. In an embodiment, the present disclosure combines CBC / Diff results from a Siemens Healthineers ADVIA® 2120i Hematology System with a multi-parametric machine learning algorithm for rapid flagging of sepsis positives. The generation of diagnostic data and the execution of the algorithm on the data are performed on the same instrument to ensure easy data access. Additionally, the algorithm can also be executed on a connected laboratory information system. Such an algorithm can be beneficial when access to sepsis testing is limited and for predicting sepsis status in unsuspected individuals to prompt confirmatory testing. Various pandemic and endemic applications for this algorithm are demonstrated.

[0126] Methods: Pathogen-positive and -negative CBC / Diff data were obtained from a preselected cohort. Using these data, a machine learning classification model was trained with 5-fold cross-validation to distinguish positives from negatives.

[0127] Results: The final ML model is an ensemble of decision tree-based methods with good accuracy and adequate AUROC on the test set. Key features include eosinophil and lymphocyte counts as previously reported. The model is able to predict hospitalized patients more accurately than non-hospitalized patients.

[0128] Conclusions: A predictive algorithm, method, and system for rapid sepsis flagging using data from routine CBC testing is provided. The algorithm may be useful in pandemic or endemic scenarios as an alternative to testing, reducing testing burden and flagging unsuspected positives.

[0129] Thus, in accordance with the present disclosure, there is provided a system, article of manufacture, and method of making and using the same, which fully fulfills the objects and advantages set forth hereinabove. While the present disclosure has been described in conjunction with the specific figures, experiments, results, and text set forth hereinabove, it is apparent that many alternatives, modifications, and variations will be apparent to those skilled in the art. Accordingly, it is intended to embrace all such alternatives, modifications, and variations that fall within the spirit and broad scope of the present disclosure.

[0130] All references cited herein are hereby incorporated by reference in their entirety.

Claims

1. A system (100) for predicting the presence of a disease in a subject, the system (100) comprising: One or more processing units (101); a medical database (112) coupled to said one or more processing units (101), said medical database (112) comprising patient data; receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with the trained machine learning model; using the trained machine learning model to predict the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject; Outputting the prediction to an output unit (105). A disease prediction module (110) configured as described above; A system (100) comprising:

2. The system of claim 1 , wherein the thresholds for sensitivity and / or specificity associated with the trained machine learning model are determined based on a requirements scenario associated with a healthcare provider.

3. 3. The system of claim 2, wherein the requirements scenario associated with the healthcare provider includes at least one of reducing infectious disease testing burden or RT-PCR testing burden, substituting infectious disease testing burden, substituting RT-PCR testing, and / or determining the need for testing of a subject.

4. The system of claim 1 , wherein the plurality of parameters associated with the subject comprises a red blood cell (RBC) parameter, a platelet parameter, and a white blood cell (WBC) parameter.

5. 5. The system of claim 4, wherein the RBC parameters include one or more of hemoglobin level, hematocrit, RBC size, hemoglobin level in the individual's RBCs, mean corpuscular volume, mean corpuscular hemoglobin concentration, normal RBCs, number of RBCs with hemoglobin concentrations ≧28 g / dL and ≦41 g / dL, number of RBCs with hemoglobin volumes ≧60 fL and ≦120 fL, hemoglobin distribution width, RBC mean corpuscular volume, RBC volume distribution width, total RBC count, hemoglobin content distribution width, and hemoglobin content average.

6. 5. The system (100) of claim 4, wherein the WBC parameters include one or more of WBC type, WBC count, WBC percentage, neutrophil to lymphocyte ratio, large unstained cell to lymphocyte ratio, number of WBCs exhibiting pseudobasophilia, and WBC maturity parameters.

7. 6. The system of claim 5, further comprising a cellular hemoglobin concentration average, a number of RBCs having a hemoglobin concentration greater than 41 g / dL, and a number of RBCs having a hemoglobin concentration less than 28 g / dL.

8. The system of claim 1 , wherein the trained machine learning model is pre-selected.

9. The system (100) of claim 1, wherein the disease prediction module (110) is further configured to use the trained machine learning model to predict stratification / severity or need for hospitalization of the subject based on the plurality of parameters associated with the subject.

10. The system (100) of claim 1, wherein the disease prediction module (110) is further configured to predict the onset of COVID-19 sequelae in the subject.

11. A method (200) for predicting the presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with the trained machine learning model; using the trained machine learning model to predict the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject; outputting said prediction to an output unit; The method (200).

12. 12. The method (200) of claim 11, wherein the thresholds for sensitivity and / or specificity associated with the trained machine learning model are determined based on a requirements scenario associated with a healthcare provider.

13. 13. The method of claim 12, wherein the requirements scenario associated with the healthcare provider includes at least one of mitigating RT-PCR or other infectious disease testing burden, substituting such testing burden, and / or determining the need for testing of a subject.

14. The method (200) of claim 11, wherein the disease is COVID-19.

15. using the trained machine learning model to predict the need for hospitalization of the subject based on the plurality of parameters associated with the subject; predicting the onset of COVID-19 sequelae in the subject; 12. The method of claim 11, further comprising:

16. 1. An article of manufacture, such as a system or component thereof, including a non-transitory computer readable medium having instructions encoded thereon, the instructions configured to cause one or more processors to execute a method including predicting the presence of a disease in a subject, the method comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with the trained machine learning model; using the trained machine learning model to predict the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject; outputting said prediction to an output unit; Manufactured goods, including.

17. 1. A non-transitory computer readable medium having instructions encoded thereon configured to cause one or more processors to execute a method for predicting the presence of a disease in a subject, the method comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and / or specificity associated with the trained machine learning model; using the trained machine learning model to predict the presence of the disease in the subject based on a particular threshold of the model and the plurality of parameters associated with the subject; outputting said prediction to an output unit; A non-transitory computer readable medium comprising:

Images