Method for analysis of cough sounds using disease signatures to diagnose respiratory diseases

1. A method for diagnosing one or more diseases of the respiratory tract for a patient, the method comprising:

processing cough sounds acquired from a patient to produce cough sound feature signals representing one or more cough sound features from the cough segments;

obtaining one or more disease signatures based on the cough sound feature signals; and

classifying the one or more disease signatures to deem the cough segments as indicative of one or more of said diseases;

wherein the obtaining of the one or more disease signatures based on the cough sound feature signals includes applying the cough sound features to each of a plurality of pre-trained disease signature decision machines, at least one of the pre-trained disease signature decision machines having been pre-trained to classify the cough sound features as corresponding to either a particular disease or to a non-disease state, and at least one other of the pre-trained disease signature decision machines having been pre-trained to classify the cough sound features as corresponding to a first particular disease or a second particular disease different from the first particular disease,

wherein a plurality of disease signatures are obtained, each said disease signature being generated as a respective axis having values varying continuously between 0 and 1 and defining a respective space for distinguishing between two different diagnoses,

wherein a response is provided for the patient along each of the axes, and

wherein the responses for the axes are provided for the classifying.

2. The method according to claim 1 , wherein the disease signature decision machines each comprise a trained Logistic Regression Model (LRM).

3. The method according to claim 2 , wherein each trained Logistic Regression Model (LRM) is trained using a reduced set of training features being features determined to be significant to the LRM to thereby avoid overtraining of the LRM.

4. The method according to claim 3 , wherein the training features are determined to be significant to the LRM by computing mean p-values for all training features and then selecting the features with mean p-values less than a threshold P min .

5. The method according to claim 4 , wherein independent variables for each LRM are values of the cough sound features and the output value from the LRM comprises a first prediction probability of the cough indicating the first particular disease with reference to the second particular disease for the at least one other of the pre-trained disease signature decision machines having been pre-trained to classify the cough sound features, and a second prediction probability with reference to the non-disease state for the at least one of the pre-trained disease signature decision machines having been pre-trained to classify the cough sound features as corresponding to either a particular disease or to a non-disease state.

6. The method according to claim 1 , wherein one or more of the plurality of pre-trained signature decision machines include one or more trained neural networks.

7. The method according to claim 1 , further comprising applying clinical patient measurements, in addition to the cough features, as independent variables to the disease signature decision machines.

8. The method according to claim 1 , wherein the classifying of the one or more disease signatures to deem the cough segments as indicative of one or more of said diseases includes applying the disease signatures to a single classifier that is trained to cover all disease groups.

9. The method according to claim 1 , wherein the classifying of the one or more disease signatures to deem the cough segments as indicative of one or more of said diseases includes applying the one or more disease signatures to a multiplicity of classifiers that are each trained to recognize a disease of interest.

10. The method according to claim 6 , wherein the classifying of the one or more disease signatures includes classifying the one or more disease signatures as one of the following diseases: Bronchiolitis (S Bo ); Croup (S C ); Asthma/RAD (S A ); Pneumonia (S P ); Lower respiratory tract diseases (S LRTD ); Primary URTI (S U ).

11. The method according to claim 9 , wherein d disease signatures are produced by the obtaining of the one or more disease signatures based on the cough sound feature signals and correspondingly the Artificial Neural Network (ANN) has a d-dimensional input layer with one input neuron corresponding to each of the disease signatures.

12. The method according to claim 11 , wherein the ANN has a k-dimensional output layer wherein each neuron in the output layer outputs a probability corresponding to a disease.

13. The method according to claim 1 , wherein the classifying of the one or more disease signatures includes compiling a compound effective probability measure P Q ′ as:

P Q ′=P Q (1− P Z )

where P Q comprises an indicator of the probability of the patient belonging to disease Q and P z to disease Z whereby the product P Q (1−P Z ) indicates the probability of the compound event that the patient belongs to disease Q AND does not belong to disease Z.

14. The method according to claim 10 , further comprising computing a cough index for each of a plurality of target diseases, wherein the cough index of a patient for a given target disease is computed as the ratio of coughs of the patient classified as indicative of the given target disease to the total number of coughs analyzed for said patient.

15. The method according to claim 1 , further comprising applying one or more post screening tests to classifications from classifying of the one or more disease signatures to detect dominant false positives.

16. The method according to claim 15 , further comprising adjusting the classifying of the one or more disease signatures based on the detected false positives.

17. The method according to claim 1 , further comprising applying specific therapy to the patient based upon the particular disease diagnosed.

18. The method according to claim 1 , wherein at least some of the axes embed domain-specific knowledge, and

wherein a proper subset of signatures are used in the classifying, based on a disease group being diagnosed.

19. A method for diagnosing one or more diseases of the respiratory tract for a patient, the method comprising:

acquiring cough sounds from the patient;

processing the cough sounds to produce cough sound feature signals representing one or more cough sound features from the cough segments;

obtaining one or more disease signatures based on the cough sound feature signals;

classifying the one or more disease signatures to deem the cough segments as indicative of one or more of said diseases; and

applying one or more post screening tests to classifications from classifying of the one or more disease signatures to detect dominant false positives,

wherein the obtaining of the one or more disease signatures based on the cough sound feature signals includes applying the cough sound features to each of one or more pre-trained disease signature decision machines, each said decision machine having been pre-trained to classify the cough sound features as corresponding to either a particular disease or to a non-disease state as corresponding to first particular disease or a second particular disease different from the first particular disease, and

wherein the one or more disease signature decision machines each comprise a trained Logistic Regression Model (LRM).

20. The method according to claim 19 , wherein each trained LRM is trained using a reduced set of training features being features determined to be significant to the LRM to thereby avoid overtraining of the LRM.

21. The method according to claim 20 , wherein the training features are determined to be significant to the LRM by computing mean p-values for all training features and then selecting the features with mean p-values less than a threshold P min .

22. The method according to claim 21 , wherein independent variables for each LRM are values of the cough sound features and the output value from the LRM comprises a first prediction probability of the cough indicating the first particular disease with reference to the second particular disease for the at least one other of the pre-trained disease signature decision machines having been pre-trained to classify the cough sound features, and a second prediction probability with reference to the non-disease state for the at least one of the pre-trained disease signature decision machines having been pre-trained to classify the cough sound features as corresponding to either a particular disease or to a non-disease state.

23. The method according to claim 19 , further comprising applying clinical patient measurements, in addition to the cough features, as independent variables to the disease signature decision machines.

24. The method according to claim 19 , wherein the classifying of the one or more disease signatures to deem the cough segments as indicative of one or more of said diseases includes applying the disease signatures to a single classifier that is trained to cover all disease groups.

25. The method according to claim 19 , wherein the classifying of the one or more disease signatures to deem the cough segments as indicative of one or more of said diseases includes applying the one or more disease signatures to a multiplicity of classifiers that are each trained to recognize a disease of interest.

26. The method according to claim 19 , wherein the classifying of the one or more disease signatures includes classifying the one or more disease signatures as one of the following diseases: Bronchiolitis (S Bo ); Croup (S C ); Asthma/RAD (S A ); Pneumonia (S P ); Lower respiratory tract diseases (S LRTD ); Primary URTI (S U ).

27. The method according to claim 19 , further comprising adjusting the classifying of the one or more disease signatures based on the detected false positives.

28. The method according to claim 19 , further comprising applying specific therapy to the patient based upon the particular disease diagnosed.