Search

EP-4740214-A1 - SYSTEMS AND METHODS FOR AI BASED INFECTION CLASSIFICATION

EP4740214A1EP 4740214 A1EP4740214 A1EP 4740214A1EP-4740214-A1

Abstract

Methods, systems, and computer storage media may be used to detect and differentiate bacterial and viral infections. Cellular population parameters such as the monocyte distribution width (MDW) can be acquired for an individual. Clinical measurements such as a complete blood count (CBC), a measure of procalcitonin (PCT) or a measure of C-reactive protein (CRP) can be obtained for an individual. A trained model can be used to compute an output value from the parameters. The output value can be compared to a threshold value to determine a probability associated with a bacterial or viral infection. A report can be generated from the probability and sent to the appropriate party.

Inventors

  • LIAO, Jiexi
  • TSALIK, EPHRAIM
  • PERSING, Dave

Assignees

  • Cepheid

Dates

Publication Date
20260513
Application Date
20240703

Claims (20)

  1. 1. A method of classifying an infection comprising: analyzing a blood sample drawn from an individual for at least a first cell population parameter for the individual; determining, for the individual, at least a second cell population parameter, wherein the second cell population parameter is selected from eosinophil count, neutrophil count, monocyte distribution width (MDW), red blood cell (RBC) count, basophil count, lymphocyte count, a measure of hemoglobin, white blood cell (WBC) count, red cell distribution width (RDW), monocyte count, a measure of hematocrit level, a measure of mean platelet volume (MPV), platelet count, ratios thereof, or a combination thereof; generating an input data set for the individual including the first cell population parameter and the second cell population parameter; providing the input data set as input to a trained model configured to generate an output value; and responsive to the output value of the trained model exceeding a predetermined threshold, communicating an indication of a prediction probability favoring a bacterial infection of the individual.
  2. 2. The method of claim 1, wherein analyzing the blood sample drawn from an individual for at least a first cell population parameter comprises analyzing a monocyte cell population parameter, a granulocyte cell population parameter, a lymphocyte cell population parameter, a red blood cell population parameter, white blood cell count, platelet count, mean platelet volume (MPV), ratios thereof, or a combination thereof.
  3. 3. The method of claim 1 or 2, wherein the method comprises analyzing a monocyte cell population parameter.
  4. 4. The method of claim 3, wherein the monocyte cell population parameter comprises monocyte distribution width (MDW), monocyte count, standard deviation in monocyte volume, mean monocyte volume, reactive monocytes, ratios thereof, or a combination thereof.
  5. 5. The method of any one of claim 2 to claim 4, wherein the method comprises analyzing the granulocyte cell population parameter.
  6. 6. The method of claim 5, wherein the granulocyte cell population parameter comprises eosinophil count, neutrophil count, basophil count, neutrophil volume distribution width, standard deviation in the volume of neutrophils, ratios thereof, or a combination thereof.
  7. 7. The method of any one of claim 2 to claim 6, wherein the method comprises analyzing the red blood cell population parameter.
  8. 8. The method of claim 7, wherein analyzing the red blood cell population parameter comprises red blood cell count, red cell distribution width (RDW), a measure of hemoglobin, a measure of hematocrit level, ratios thereof, or a combination thereof.
  9. 9. The method of any one of claim 2 to claim 8, wherein the method comprises analyzing the lymphocyte cell population parameter.
  10. 10. The method of claim 9, wherein analyzing the lymphocyte cell population parameter comprises lymphocyte count, standard deviation in lymphocyte volume, mean lymphocyte volume, reactive lymphocyte, ratios thereof, or a combination thereof.
  11. 11. The method of any one of claim 2 to claim 10, wherein the method comprises analyzing a neutrophil to lymphocyte ratio (NLR), eosinophil to lymphocyte ratio (ELR), lymphocyte to monocyte ratio (LMR), eosinophil-to-monocyte ratio (EMR), platelet-to-lymphocyte ratio (PLR), platelet-to-neutrophil ratio, monocyte to lymphocyte ratio (MLR), mean platelet volume-to-platelet count (MPV/PC) ratio, basophil to lymphocyte (BLR) ratio, or combinations thereof.
  12. 12. The method of claim 11, wherein the method comprises analyzing a neutrophil to lymphocyte ratio (NLR).
  13. 13. The method of any one of claim 1 to claim 12, wherein the at least second cell population parameter includes eosinophil count, neutrophil count, or a combination thereof.
  14. 14. The method of any one of claim 1 to claim 13, wherein the input data set for the individual includes red blood cell (RBC) count, basophil count, lymphocyte count, a measure of hemoglobin, white blood cell (WBC) count, red cell distribution width (RDW), monocyte count, a measure of hematocrit level, a measure of mean platelet volume (MPV), platelet count, or ratios thereof.
  15. 15. The method of any one of claim 1 to claim 14, wherein the input data set for the individual includes eosinophil count, neutrophil count, monocyte distribution width (MDW), red blood cell (RBC) count, basophil count, lymphocyte count, a measure of hemoglobin, white blood cell (WBC) count, red cell distribution width (RDW), monocyte count, a measure of hematocrit level, a measure of mean platelet volume (MPV), platelet count, and neutrophil to lymphocyte ratio.
  16. 16. The method of any one of claim 1 to claim 13, wherein the input data set for the individual additionally includes at least one clinical measurement selected from a complete blood count (CBC), erythrocyte sedimentation rate (ESR), anion gap, a measure of procalcitonin (PCT), a measure of C-reactive protein (CRP), alkaline phosphatase (ALK), aspartate aminotransferase (AST), calcium, oxygen (O2) saturation, or ratios thereof.
  17. 17. The method of any one of claim 1 to claim 16, further comprising: communicatively coupling with a database; and retrieving, from the database, the second cell population parameter, the clinical measurement, or combinations thereof, for the individual.
  18. 18. A method of classifying an infection comprising: analyzing a blood sample drawn from an individual for a monocyte distribution width (MDW) value for the individual; determining, for the individual, an eosinophil count and neutrophil count; generating an input data set for the individual including the MDW value, the eosinophil count, and the neutrophil count; providing the input data set as input to a trained model configured to generate an output value; and responsive to the output value of the trained model exceeding a predetermined threshold, communicating an indication of a prediction probability favoring a bacterial infection of the individual.
  19. 19. The method of any one of claim 1 to claim 18, wherein the individual is a human.
  20. 20. The method of any one of claim 1 to claim 19, wherein the at least second cell population parameter, a molecular parameter, the clinical measurement, or combinations thereof, are determined from the blood sample.

Description

SYSTEMS AND METHODS FOR Al BASED INFECTION CLASSIFICATION BACKGROUND CROSS-REFERENCE TO RELATED APPLICATIONS [0001] This application is an international application which claims priority to and benefit of U.S. Provisional Application Serial No. 63/525,554 filed July 7, 2023, the contents of which are incorporated by reference in their entirety for all purposes. [0002] Erroneous prescription of antibacterials to patients poses a significant issue within the medical community. Antibacterial prescription can lead to loss and a disruption of the gut microflora which can result in opportunistic bacterial infections, such as Clostridioides difficile which causes painful inflammation and potentially can be life-threatening. Additionally, unnecessary exposure can also lead to antimicrobial resistance within bacterial populations. Nevertheless, the proper use of antibacterials in situations where a bacterial infection is the root cause of problematic symptoms can be lifesaving. [0003] Often, symptoms relating to a bacterial infection or viral infection can be very similar to one another. In such situations, it can become difficult for a medical professional to differentiate between a viral versus a bacterial infection. Current approaches utilize procalcitonin (PCT), c-reactive protein (CRP), white blood cell (WBC), and neutrophil counts in a clinical setting to make early diagnosis of bacterial infections. However, studies suggest that these parameters do not reliably differentiate between a bacterial versus viral infection; meta-analysis of PCT in pneumonia found a 55% sensitivity and 76% specificity with AUROCs that vary from 0.60-0.85 (see PMID 31241140). In such cases, a medical professional may erroneously prescribe an antibacterial when a viral infection is actually the root cause of the patient’s symptoms. Conversely, antibacterials may not be timely administered when a bacterial infection is mistakenly thought to be a viral infection. Consequently, systems and methods that more accurately aid medical professionals in differentiating between bacterial and viral infections are desirable within the medical community. BRIEF SUMMARY [0004] In various aspects, systems and methods are provided for classifying an infection for an individual as more likely to be bacterial or more likely to be viral, based on a prediction probability. The prediction probability can be derived by measuring and inputting cell population parameters of an individual into a trained model. In an aspect, cell population parameters used as an input to the trained model includes one or more of the following: a monocyte distribution width (MDW), a eosinophil count, neutrophil count, monocyte distribution width (MDW), red blood cell (RBC) count, basophil count, lymphocyte count, a measure of hemoglobin, white blood cell (WBC) count, red cell distribution width (RDW), monocyte count, a measure of hematocrit level, a measure of mean platelet volume (MPV), and a platelet count, a neutrophil to lymphocyte ratio (NLR), eosinophil to lymphocyte ratio (ELR), lymphocyte to monocyte ratio (LMR), eosinophil-to-monocyte ratio (EMR), platelet-to-lymphocyte ratio (PLR), platelet-to-neutrophil ratio, monocyte to lymphocyte ratio (MLR), mean platelet volume-to-platelet count (MPV/PC) ratio, basophil to lymphocyte (BLR) ratio and any other measurements that measure and/or compares the number, types, and/or properties of cells for an individual. In response to this input, an output value is generated. The output value may be compared to a threshold level to determine a prediction probability of a bacterial or viral infection. In other words, the cell population parameters, through the use of a trained model, can be used to determine a predictive likelihood of a bacterial infection over a viral infection or vice versa. [0005] In various aspects, a trained model is generated using a population of individuals with known bacterial or viral infections, of which cell population parameters have been measured. The training model uses cell population parameters as an input which is processed through a model program to provide output data. The output data, when compared to a predicted output, is used to access the model. Through multiple rounds of iteration, the model is optimized until an acceptable level of predictability is obtained. The model can then be applied to cell population parameters from an individual where a bacteria or viral infection is in question and a prediction probability can be obtained. [0006J In various aspects, cell population parameters and other physiological parameters, such as clinical measurements and molecular parameters, or patient demographic parameters, can be combined for an individual to determine a prediction probability. The prediction probability is then derived by inputting the combined physiological parameters and optionally patient demographic parameters into a trained model that has been optimized for all parameters being eva