Screening of individuals for a respiratory disease using artificial intelligence

Abstract

An artificial intelligence-based system and method for scalable screening of individuals for respiratory infection, such as COVID-19. The system is trained to distinguish distinct latent features of cough sounds produced by a COVID-19 infected person from cough sounds produced by patients suffering from any other respiratory infection or involuntary cough sounds produced by a healthy person. Cough sound samples from individuals can be remotely collected and evaluated by the system for likelihood of the COVID-19 infection. Additionally, images of affected body parts, biomarkers, metadata, and other respiratory sound samples can also be used for screening.

What is claimed is: 
     
         1 . A method for predicting a likelihood of having a respiratory infection in an individual, the method implemented within a screening system, the screening system has a processor and a memory, the method comprising the steps of:
 receiving, by the screening system, cough samples from subjects suffering from the respiratory infection and subjects not suffering from the respiratory infection;   pre-processing the cough samples, by a pre-processing module implemented within the screening system, into labeled features obtaining a training dataset;   randomly partitioning the training dataset into a plurality of training datasets;   generating a plurality of classifiers, by the screening system using machine learning, based on the plurality of training datasets, wherein at least one classifier of the plurality of classifiers is a deep learning classifier; and   assigning a weightage, by the screening system, to each classifier of the plurality of classifiers.   
     
     
         2 . The method according to  claim 1 , wherein the method further comprises the steps of:
 receiving, by the screening system, a sound sample from an individual, the sound sample includes a cough sound;   pre-processing the sound sample, by the pre-processing module, to remove noise and extract a set of unique latent features;   processing the set of unique latent features by each of the plurality of classifiers to obtain a plurality of outputs, each output of the plurality of outputs indicates a likelihood of having the respiratory infection or not having the respiratory disease;   computing, by a mediator module implemented within the system, an average weightage of the plurality of outputs based on the weightage assigned to each classifier of the plurality of classifiers to obtain a result, wherein the result classifies the individual into a category indicative of a likelihood of having the respiratory infection or into another category indicative of a likelihood of not having the respiratory infection; and   communicating, by the mediator module, the result to the individual.   
     
     
         3 . The method according to  claim 1 , wherein the method further comprises the steps of:
 configuring, the pre-processing module and the plurality of classifiers into an individual device;   receiving a sound sample from an individual through a microphone coupled to the individual device, the sound sample includes a cough sound;   pre-processing the sound sample, by the pre-processing module, to remove noise and extract a set of unique latent features;   processing the set of unique latent features by each of the plurality of classifiers to obtain a plurality of outputs, each output of the plurality of outputs indicates a likelihood of having the respiratory infection or not having the respiratory infection;   computing, by the individual device, an average weightage of the plurality of outputs based on the weightage assigned to each classifier of the plurality of classifiers to obtain a result, wherein the result classifies the individual into a category indicative of a likelihood of having the respiratory infection or into another category indicative of a likelihood of not having the respiratory infection; and   presenting the result on a display coupled to the individual device.   
     
     
         4 . The method according to  claim 1 , wherein the respiratory infection is COVID-19 infection. 
     
     
         5 . The method according to  claim 2 , wherein the method further comprises the steps of:
 computing a plurality of results for a plurality of individuals;   receiving a plurality of location coordinates of the plurality of individuals; and   applying machine learning algorithms to the plurality of individuals and the plurality of location coordinates for presenting in near real time rate and pattern of a spread of the respiratory infection.   
     
     
         6 . The method according to  claim 5 , wherein the method further comprises the steps of:
 predicting, by the screening system, cluster spreads of the respiratory infection in near future.   
     
     
         7 . The method according to  claim 6 , wherein the method further comprises the steps of receiving past mobility patterns, from an external server, of the plurality of individuals, wherein the predicting of the cluster spreads is further based upon the past mobility patterns. 
     
     
         8 . The method according to  claim 1 , wherein the method further comprises the steps of receiving, by the screening system, a plurality of parameters selected from a group consisting of biomarkers, metadata, neurologic manifestations including central nervous system manifestations and peripheral nervous system manifestations, and image(s) of cutaneous manifestations of the respiratory infection in a form of anomalies in skin; and
 pre-processing the plurality of parameters to obtain a plurality of additional training datasets, wherein the generating of the plurality of classifiers is further based on the plurality of additional training datasets.   
     
     
         9 . The method according to  claim 8 , wherein the biomarkers are selected from a group consisting of respiration rate, body temperature, blood oxygen saturation, pulse rate, heart rate variability, resting heart rate, blood pressure, mean arterial pressure, stroke volume, sweat level, systematic vesicular resistance, cardiac output, pulse pressure, cardiac index, one lead ECG, and breath chemical composition including volatile organic compounds. 
     
     
         10 . The method according to  claim 9 , wherein the metadata is selected from a group consisting of age, gender, smoking, non-smoking, ethnicity, and medical history. 
     
     
         11 . The method according to  claim 9 , wherein the central nervous system manifestations are selected from a group consisting of dizziness, headache, impaired consciousness, acute cerebrovascular disease, ataxia, and seizure, and the peripheral nervous system manifestations are selected from a group consisting of taste impairment, smell impairment, vision impairment, and nerve pain. 
     
     
         12 . The method according to  claim 8 , wherein the step of assigning the weightage to each classifier is further based on a parameter of the plurality of parameters used for generating each classifier. 
     
     
         13 . A method for predicting a likelihood of having a respiratory infection in an individual, the method implemented within a screening system, the screening system has a processor and a memory, the method comprising the steps of:
 receiving, by the screening system, a sound sample from an individual, the sound sample includes a cough sound;   pre-processing the sound sample, by a pre-processing module, to remove noise and amplification, the pre-processing module implemented within the screening system;   applying a deep learning-based classifier to detect cough sound in the sound sample;   applying a feature extraction algorithm to the sound sample to extract a set of unique latent features; and   processing the set of unique latent features by one or more machine learning based classifiers to classify the individual into a category indicative of a likelihood of having the respiratory infection or into another category indicative of a likelihood of not having the respiratory infection.   
     
     
         14 . The method according to  claim 13 , wherein the respiratory infection is COVID-19 infection. 
     
     
         15 . The method according to  claim 13 , wherein the sound sample further includes nasal breathing sounds and vocalization sounds. 
     
     
         16 . The method according to  claim 13 , wherein the one or more machine learning based classifiers include a deep transfer learning-based binary classifier.