WO2021245657A1

WO2021245657A1 - System for detecting viral infections

Info

Publication number: WO2021245657A1
Application number: PCT/IL2021/050642
Authority: WO
Inventors: Haim SIBONI; Levy ZRUYA
Original assignee: Foresight Automotive Ltd.
Priority date: 2020-06-01
Filing date: 2021-05-31
Publication date: 2021-12-09

Abstract

A system for detecting viral infections which comprises a plurality of electronic sampling devices, each of which configured to non-invasively acquire a testee- specific sample related to a different symptom of a given viral infection; a classifier programmed with an algorithm for defining symptoms of the given viral infection and for prioritizing each of the symptoms in terms of their likelihood of being indicative that the testee is infected with the given virus; a trained machine learning module into which each of the testee-specific samples is input and from which symptom related features are extracted in a testing phase; a database in which a plurality of symptom related features that have been extracted in a training phase are stored. The classifier is configured to compare the testing phase extracted features with the training phase extracted features for a same symptom and to assign a symptom-specific score according to a predetermined degree of similarity between the testing phase extracted features and the training phase extracted features, wherein the classifier is additionally configured to assign a symptom- specific score for all other symptoms defined by the algorithm and to determine that the testee is infected with the given viral infection when an accumulated score which is equal to a sum of all the assigned symptom-specific scores is greater than a predetermined value.

Description

SYSTEM FOR DETECTING VIRAL INFECTIONS

Field of the Invention

The present invention relates to the field of health care apparatus. More particularly, the present invention relates to a system for detecting viral infections, especially the Corona Virus Disease 2019 (COVID-19).

Background of the Invention

The Corona virus (COVID-19) is a highly contagious virus. The Corona virus has spread rapidly around the world, and is characterized by symptoms that include fever, cough, rash, red eyes and difficulty in breathing, and in severe cases also causes acute pneumonia requiring artificial respiration and even death. Millions of patients worldwide have died from the corona virus infection.

Usually, as mentioned above, people who have been infected with the Corona virus develop four typical symptoms: increased body heat, red eyes, an unusual cough and a skin rash, such as at the back of the hands and feet. Some of these symptoms can also individually indicate the occurrence of other viral infections or of other diseases such as: a cold, laryngitis, influenza (non-corona), and skin disease. Therefore, there is a need to conclusively determine that an individual has been infected with the Corona virus.

Large scale detection of a virus in general and the Corona virus in particular is considered to be one of the most important needs for controlling and eventually eliminating a pandemic by screening healthy people from infected individuals.

The most commonly used method to test whether an individual has been infected with a virus is the mouth or nose swab antibody-detection polymerase chain reaction (PCR) test, whereby genetic material from a sample is isolated and is caused to fluoresce to indicate that it contains the virus after undergoing a forced reaction. Although the PCR test is substantially accurate, it is associated with some significant disadvantages. Firstly, the PCR test is invasive, being bothersome to the testee when the swab is introduced to the nose or throat to obtain the sample, and also requires that the sample be carefully maintained in a sterile and a low-temperature environment. Secondly, the process of sample taking, transportation of the sample to a laboratory and waiting until the laboratory results are available is prolonged, being more than 6 hours and often a few days. This prolonged time can be critical when dealing with mass tests of a large population in case of a widespread pandemic.

Antibody tests carried out with a blood test are also problematic during a pandemic, since it can take five days after the initial infection for the human body to produce enough antibodies to be detected with the corresponding test-kits. To make matters worse, it can take hours, or even days, to obtain the results from these tests due to the long queue. At the same time, human carriers continue to spread the virus and be contagious.

It is an object of the present invention to provide a system for speedily and accurately detecting people who have become infected with a specific virus, and particularly the Corona virus (COVID-19).

It is another object of the present invention to provide a system for non-invasively testing people who are suspect of being infected with a specific virus.

It is another object of the present invention to provide a system for at-home testing of people who are suspect of being infected with a specific virus, such as the Corona virus.

Other objects and advantages of the invention will become apparent as the description proceeds. Summary of the Invention

A system for detecting viral infections comprises a plurality of electronic sampling devices, each of which configured to non-invasively acquire a testee-specific sample related to a different symptom of a given viral infection; a classifier programmed with an algorithm for defining symptoms of the given viral infection and for prioritizing each of the symptoms in terms of their likelihood of being indicative that the testee is infected with the given virus; a trained machine learning module into which each of the testee-specific samples is input and from which symptom related features are extracted in a testing phase; and a database in which a plurality of symptom related features that have been extracted in a training phase are stored, wherein the classifier is configured to compare the testing phase extracted features with the training phase extracted features for a same symptom and to assign a symptom-specific score according to a predetermined degree of similarity between the testing phase extracted features and the training phase extracted features, wherein the classifier is additionally configured to assign a symptom-specific score for all other symptoms defined by the algorithm and to determine that the testee is infected with the given viral infection when an accumulated score which is equal to a sum of all the assigned symptom-specific scores is greater than a predetermined value.

As referred to herein, a "sample" is a data file representative of a symptom of the given viral infection and associated with a specific time window.

In one aspect, the system further comprises a server provided with the classifier, trained machine learning module and database.

In one aspect, the system further comprises a terminal device comprising each of the sampling devices.

In one aspect, each of the sampling devices is housed in a monolithic casing of the terminal device. In one aspect, the terminal device is deployed in a public place.

In one aspect, the terminal device is an at-home automated tool.

In one aspect, the classifier is operable to prioritize one or more of the symptoms such that a test involving the system will be discontinued when an assigned individual or accumulated symptom-specific score for the one or more systems is less than another predetermined value to indicate that the testee is not infected with the given viral infection.

In one aspect, the classifier is programmed with an algorithm for defining symptoms of COVID-19, such as one that prioritizes the symptoms of a predetermined high body temperature and a characteristic cough.

The sampling devices for acquiring a testee-specific sample related to a predetermined high body temperature and a characteristic cough, respectively, may be a camera with an infrared detector and a microphone having sound analysis and environmental noise filtering abilities. The infrared detector is preferably sensitive in a longwave infrared (LWIR) range which is suitable for measuring human body temperature

In one aspect, another one of the sampling devices is a camera which is sensitive to visible light, for skin texture analysis to indicate if a skin rash is detected and for detecting eye redness when a captured image is magnified.

In one aspect, another one of the sampling devices is an odor generator to test for olfactory impairment.

A method for detecting viral infections comprises non-invasively acquiring in a testing phase one or more testee-specific samples related to a different symptom of a given viral infection with corresponding electronic sampling devices, extracting features from said acquired samples using a trained machine learning module, comparing the testing phase extracted features with training phase extracted features for a same symptom, and determining that a testee is infected with the given viral infection when the testing phase extracted features and the training phase extracted features for a same symptom have at least a predetermined degree of similarity.

Brief Description of the Drawings

In the drawings:

Fig. 1 is a schematic illustration of an embodiment of a viral infection detecting system;

Fig. 2 is a schematic illustration of a terminal device usable in conjunction with the system of Fig. 1;

Fig. 3 is a method for determining whether a testee is infected with COVID- 19; and

Fig. 4 is a method for determining whether a person is infected with a given viral disease, during both training and testing phases.

Detailed Description of the Invention

A person infected with a certain viral disease exhibits various known symptoms. To determine whether a given testee exhibits these known symptoms, the viral infection detecting system, which may be a user-friendly, at-home automated tool, or alternatively deployed at a public site, comprises a plurality of sampling devices, each of which configured to acquire testee-specific data related to a different symptom. The acquired data is compared with features previously extracted by a trained machine learning module that are known to be representative of the given symptom. If the acquired testee-specific data is found to be similar to one of the stored features, the system derives a relatively high symptom similarity score for the testee. This procedure is repeated for each of the known symptoms. The system automatically determines that the testee has a strong liklihood to be infected with the virus when the accumulated score for all of the symptoms is greater than a predetermined value.

The following description relates to the positive detection of people who are infected with the Corona virus (COVID-19), even though other viral infections may have some similar symptoms. It will be appreciated that the system of the invention is also suitable to detect people who are infected with other viral diseases as well, mutatis mutandis.

Fig. 1 schematically illustrates one embodiment of a viral infection detecting system 10. System 10 comprises a remote server 5 which is configured with a trained machine learning module 6 and with at least one classifier 9. Machine learning module 6 has a database 7 in which are stored various features that are known to be representative of each symptom of the viral infection being tested. A local terminal device 12 in data communication with server 5 comprises a plurality of sampling devices, e.g. sampling devices 14a-c, each of which adapted to non-invasively acquire a different symptom related sample that is derived from a testee. Terminal device 12 transmits the samples to server 5 via signal L. A controlled software based unit of server 5 receives signal L and transfers it to machine learning module 6. Classifier 9 compares features output from machine learning module 6 with the features stored in database 7 according to predetermined instructions and assigns a symptom-specific score that is indicative of whether the testee is infected with the virus being tested. Server 5 may also be in data communication with a server 15 of a health organization, for the purpose of informing the health organization that the testee has been infected, if necessary precautions have to be taken.

It will be appreciated that server 5 may have more than one machine learning module 6, database 7 and classifier 9, or alternatively a first classifier may be substituted with a second classifier, in order to detect different types of viral infections. Terminal device 12 may be used in the comfort of one's home, or may be deployed at any suitable public site such as a mall, office, school, factory, airport, and stadium.

In other embodiments, server 5 is suitable to detect more than one viral infection. Each classifier 9 is programmed with an algorithm associated with a different viral disease and cooperates with a corresponding machine learning module 6 that is trained with samples of a corresponding viral disease.

In other embodiments, one or both of database 7 and classifier 9 are copied into the memory of the processor of terminal device 12. Terminal device 12 in turn will update server 5 as to whether the testee has been found to be infected with the virus.

Classifier 9 may be configured to assign the same weight to each symptom, or alternatively may assign a higher weight to the data associated with selected symptoms or filter out a test if the acquired data related to one of the symptoms is completely dissimilar to the stored features.

Fig. 2 schematically illustrates a terminal device 12 used for detecting whether a person has become infected with the Corona virus.

A person known to be infected with COVID-19 may exhibit the following four characteristic symptoms: a. High body temperature b. Unusual cough c. Hand skin rash d. Conjunctivitis, or eye redness

It should be noted that these symptoms generally appear simultaneously when a person is infected with COVID-19; otherwise, the appearance of one or more of these symptoms may be indicative of another disease, such as the common cold or a skin disease. On the other hand, some people infected with COVID-19 have only some of these symptoms at the same time, and develop the other symptoms at a later stage. Accordingly, accurate matching of acquired samples of these symptoms with stored features will assist in properly detecting people infected with COVID-19.

In order to obtain the necessary samples for determining whether the testee is infected with COVID-19, terminal device 12 comprises the following sampling devices:

(1) Sampling device 14a embodied as a camera that is sensitive to infrared radiation, particularly in the longwave infrared (LWIR) range which is suitable for measuring human body temperature, of course after using a blackbody device for calibration purposes.

(2) Sampling device 14b embodied as a microphone, with a speech processing function, for recording the testee when coughing and filtering other environmental noise. The typical cough of a person infected with COVID-19 is significantly different from a regular cough. The speech processing function generally employs Fast Fourier Transform (FFT) to convert the recorded audio signal into individual spectral components, so that the frequency and magnitude of each spectral component can be analyzed.

(3) Sampling device 14c embodied as a camera sensitive to visible light, for skin texture analysis to indicate if a skin rash is detected. When the captured image is suitably magnified, sampling device 14c will be able to detect eye redness.

Sampling devices 14a-c are preferably housed in a compact, monolithic casing that facilitates each sample to be effortlessly acquired upon suitable positioning of the casing or entry of a selected setting, usually in response to predetermined instructions. If so desired, each sampling devices 14a-c may be housed in a separate casing. Terminal device 12 also comprises transducer circuitry and software 16 for converting the acquired sample to digital form so that it could be compared with a stored feature. For example, one or more of the features of pattern, texture, color and contrast that are able to be extracted from image data can be analyzed. A processor 17 processes the acquired data when it is compared locally with a stored feature. Communication apparatus 18 is provided to transmit the acquired data to the remote server when the comparison is performed remotely.

According to another embodiment, terminal device 12 additionally comprises the following sampling device to more accurately classify the testees:

(4) Sampling device 19 comprises an odor generator by which a predefined odor is emitted to a close vicinity of the testee. Terminal device 12 tests the testee's reaction to the emitted odor, such as visually or electronically, since a secondary symptom of COVID-19 is olfactory impairment. Therefore, an inability of the testee to smell the emitted odor may be indicative that the testee has been infected by the Corona virus.

Fig. 3 schematically illustrates an algorithm for determining whether the testee is infected with COVID-19.

In step 20, the classifier is programmed with an algorithm for defining the symptoms of a given viral infection and for prioritizing each of the symptoms in terms of their likelihood of being indicative of being infected with the given virus.

The testee is first tested for a high body temperature in step 22 by using sampling device 14a and comparing the measured temperature with a nominal temperature stored in memory. The testee is given a first score in step 24 if the difference between the measured temperature and the nominal temperature is greater than a first predetermined value. Afterwards, the testee is tested for an unusual cough in step 26 by using sampling device 14b and extracting voice related features. The testee is given a second score in step 28 if one or more of the extracted voice related features, for example frequency components and magnitude, is similar to any of the stored voice related features in accordance with stored criteria. The value of the second score is set according to the degree of similarity, with the value of the second score being higher if an extracted voice related feature is more similar to a stored voice related feature. The degree of similarity may be related to the number of stored features to which the extracted feature is similar.

If the sum of the first and second scores is less than a second predetermined value, the testee is determined to be not infected with COVID-19 and the test is discontinued in step 30. However, if the sum of the first and second scores is greater than the second predetermined value, the test continues to step 32.

In step 32, the testee is tested for a hand rash using sampling device 14c. Features extracted from the image data are compared with stored hand rash related features, The testee is given a third score in step 34 if one or more of the extracted hand rash related features is similar to any of the stored rash related features in accordance with stored criteria, with the value of the third score being set according to the degree of similarity. The testee is then tested for eye redness in step 36 after the user increases the magnification and aims the sampling device at the eyes, and is given a fourth score in step 38 if one or more of the extracted eye redness related features is similar to any of the stored eye redness related features in accordance with stored criteria, with the value of the fourth score being set according to the degree of similarity.

If the accumulated score of the first, second, third and fourth scores is greater than a third predetermined value, the classifier determines in step 40 that the testee is infected with COVID-19.

This method may similarly be performed a test using sampling device 19 is incorporated. In order to be able to compare newly extracted features with stored features and to reliably determine whether the newly extracted features are reflective of the given viral disease, the stored features have had to previously undergo training.

Fig. 4 illustrates a method for determining whether a person is infected with a given viral disease, during both training and testing phases.

Samples are taken from people who are known to be infected with the given viral disease in step 46. A machine learning module, for example based on a neural network model, is built in step 48, whereby features are extracted from the samples during a training phase in step 50 that often involves many iterations and tuning of hyperparameters as well known to those skilled in the art so that the samples will be suitably characterized, such as by patterns that are common to all the samples. These extracted features are stored in a database of symptom related features in step 52.

In step 54, samples are subsequently taken from people in a testing phase who are suspected of being infected with the given viral disease. These samples are input to the trained machine learning module in step 56. Symptom related features are output in step 58 and are compared by the classifier with corresponding stored features in step 60. The testee is then classified as being infected in step 62 when the output symptom related features are found to be substantially similar to corresponding stored features in accordance with the symptom defining and prioritizing algorithm.

As can be appreciated from the foregoing description, the system is advantageously able to speedily, automatically and reliably determine within a few minutes whether a testee is infected with a given viral disease, as opposed to the prior art which at best is able to determine infection within six hours, and generally within a few days. If for some reason, an incorrect classification was determined, such as when a not infected classification was made and the testee subsequently contracted the viral disease, the system is able to input the samples taken from that testee during the testing phase into the untrained machine learning module in order to force adjustment of the trained machine learning module.

While some embodiments of the invention have been described by way of illustration, it will be apparent that the invention can be carried out with many modifications, variations and adaptations, and with the use of numerous equivalents or alternative solutions that are within the scope of persons skilled in the art, without exceeding the scope of the claims.

Claims

WO 2021/245657 -IB - PCT/IL2021/050642 CLAIMS

1. A system for detecting viral infections, comprising: a) a plurality of electronic sampling devices, each of which configured to non-invasively acquire a testee-specific sample related to a different symptom of a given viral infection; b) a classifier programmed with an algorithm for defining symptoms of the given viral infection and for prioritizing each of the symptoms in terms of their likelihood of being indicative that the testee is infected with the given virus; c) a trained machine learning module into which each of the testee-specific samples is input and from which symptom related features are extracted in a testing phase; and d) a database in which a plurality of symptom related features that have been extracted in a training phase are stored, wherein the classifier is configured to compare the testing phase extracted features with the training phase extracted features for a same symptom and to assign a symptom-specific score according to a predetermined degree of similarity between the testing phase extracted features and the training phase extracted features, wherein the classifier is additionally configured to assign a symptom-specific score for all other symptoms defined by the algorithm and to determine that the testee is infected with the given viral infection when an accumulated score which is equal to a sum of all the assigned symptom-specific scores is greater than a predetermined value.

2. The system according to claim 1, further comprising a server provided with the classifier, trained machine learning module and database.

3. The system according to claim 1, further comprising a terminal device comprising each of the sampling devices.

4. The system according to claim 3, wherein each of the sampling devices is housed in a monolithic casing of the terminal device.

5. The system according to claim 3, wherein the terminal device is deployed in a public place.

6. The system according to claim 3, wherein the terminal device is an at-home automated tool.

7. The system according to claim 1, wherein the classifier is operable to prioritize one or more of the symptoms such that a test involving the system will be discontinued when an assigned individual or accumulated symptom-specific score for the one or more systems is less than another predetermined value to indicate that the testee is not infected with the given viral infection.

8. The system according to claim 7, wherein the classifier is programmed with an algorithm for defining symptoms of COVID-19.

9. The system according to claim 8, wherein the classifier is programmed with an algorithm for prioritizing the symptoms of a predetermined high body temperature and a characteristic cough.

10. The system according to claim 9, wherein the sampling devices for acquiring a testee-specific sample related to a predetermined high body temperature and a characteristic cough, respectively, are a camera with an infrared detector and a microphone having sound analysis and environmental noise filtering abilities.

11. The system according to claim 10, wherein the infrared detector is sensitive in a longwave infrared (LWIR) range which is suitable for measuring human body temperature

12. The system according to claim 10, wherein another one of the sampling devices is a camera which is sensitive to visible light, for skin texture analysis to indicate if a skin rash is detected and for detecting eye redness when a captured image is magnified.

13. The system according to claim 10, wherein another one of the sampling devices is an odor generator to test for olfactory impairment.