Method For Detecting and Discriminating Breathing Patterns From Respiratory Signals

• US 20080177195A1
• Filed: 12/21/2005
• Published: 07/24/2008
• Est. Priority Date: 12/23/2004
• Status: Active Grant

• Current Assignee: ResMed Pty Ltd. (ResMed Incorporated)
• Original Assignee: Resmed Limited (ResMed Incorporated)
• Inventors: Armitstead, Jeffrey
• Granted Patent: US 8,066,647 B2
• Time in Patent Office: Days
• US Class Current: 600/529
• CPC Class Codes: A61B 5/087 Measuring breath flow
• CPC Class Codes: A61B 5/145 Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissuemeasuring of blood pressure or blood flow A61B5/02; non-radiation detecting or locating of foreign bodies in blood A61B5/06
• CPC Class Codes: A61B 5/4818 Sleep apnoea
• CPC Class Codes: A61B 5/7232 involving compression of the physiological signal, e.g. to extend the signal recording period
• CPC Class Codes: A61B 5/726 using Wavelet transforms
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: A61M 16/021 operated by electrical means A61M16/202 – A61M16/205 take precedence
• CPC Class Codes: A61M 2230/40 Respiratory characteristics
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

A signal representative of a patient'"'"'s respiration is split into equal length epochs. A primary feature is extracted from each epoch that acts as a compressed representation of the signal events. Statistics are applied to the primary feature to produce one or more secondary features that represent the entire epoch. Each secondary feature is grouped with one or more other features that are extracted from the entire epoch rather than selected epoch events. This grouping is the epoch pattern. The pattern is manipulated with suitable classifier algorithm to produce a probability for each class within the algorithm, that the signal may be representative of a disease state (Cheyne-Stokes, OSA etc). The epoch is assigned to the class with the highest probability. Also defined are methods of detecting Cheyne-Stokes breathing by analysing a signal to detect one or regions of hyperpnoea and if the length of a hyperpnoea exceeds a parameter, Cheyne-Stokes is present.

Claims

1. (canceled)

2. (canceled)

3. (canceled)

4. (canceled)

5. (canceled)

6. (canceled)

7. (canceled)

8. A method for diagnosing the presence of sleep disorders comprising

recording a signal representative of respiration from a patient using a logging device which includes a data-acquisition system and a memory,

processing the respiratory signal either on-board by the recording device or off-line using a computer,

dividing the signal into n epochs of equal length,

recording events consisting of an hypopnoea followed by an hyperpnoea,

detecting for each event the beginning and end points,

calculating event lengths, and

processing each hyperpnoea to derive shape features.

9. The method for diagnosing the presence of sleep disorders of claim 8, wherein initially the signal is filtered to remove noise and the baseline is zeroed.

10. The method for diagnosing the presence of sleep disorders of claim 8, wherein the signal is linearised.

11. The method for diagnosing the presence of sleep disorders of claim 8, rejecting events if they do not follow the correct sequence of hypopnoea/apnea-hyperpnoea.

12. The method for diagnosing the presence of sleep disorders of claim 8, rejecting events that fall outside predetermined length scale limits.

13. The method for diagnosing the presence of sleep disorders of claim 8, wherein the shape features are calculated using singular value decomposition of the hyperpnoea ventilation signal by the steps of

extracting the hyperpnoea from the respiratory signal,

forming a ventilation signal from the respiratory signal,

scaling the ventilation signal to give a vector of values,

calculating a set of orthogonal functions arranged as a matrix, and

calculating shape factors from the product of the pseudoinverse of the matrix and the vector of values.

14. The method for diagnosing the presence of sleep disorders of claim 13, wherein the matrix of orthogonal functions is:

15. The method for diagnosing the presence of sleep disorders of claim 8, calculating shape features as:

F_P(1-4)=V_hyperp×

PseudoInverse(M_Basis);

where the ventilation signal is scaled by its mean value to give a vector of values V_hyperp.

16. The method for diagnosing the presence of sleep disorders of claim 8, deriving a jump feature for each hyperpnoea by the steps of

extracting the hyperpnoea from the respiratory signal,

forming a ventilation signal from the absolute value of the respiratory signal, and

approximating the envelope of the ventilation signal with a droopy peak-detector,

17. The method for diagnosing the presence of sleep disorders of claim 16, wherein the following droopy peak-detector is used to approximate the envelope of the ventilation signal:

18. The method for diagnosing the presence of sleep disorders of claim 8, calculating secondary length measure features from primary features by the following equations:

19. The method for diagnosing the presence of sleep disorders of claim 8, where a secondary length measure is hypopnoea length

20. The method for diagnosing the presence of sleep disorders of claim 8, calculating the jump feature from an envelope of the signal representative of respiration.

21. The method for diagnosing the presence of sleep disorders of claim 8, calculating hyperpnoea shape features from an envelope of the signal representative of respiration:

22. The method for diagnosing the presence of sleep disorders of claim 8, calculating an additional feature from the entire epoch signal from the spectrogram of the epoch signal as follows:

subtracting the mean of the respiratory signal from the respiratory signal,

chopped the subtracted signal into n slices,

windowing each slice using a Hanning window,

applying a Fourier transform to each windowed slice yielding a complex vector result for each slice,

taking the absolute value of each complex result yielding a real valued vector per slice,

averaging the resulting vectors to yield one vector,

taking the natural log of the subsequent vector

extracting the values in the frequency range 0 Hz to 0.075 Hz to form a sub-vector de-trending the sub-vector,

calculating the feature as the maximum minus the mean of the de-trended vector.

23. The method for diagnosing the presence of sleep disorders of claim 8, deriving additional features by applying wavelet analysis to each epoch.

24. The method for diagnosing the presence of sleep disorders of claim 23, using wavelet coefficients or statistics derived from wavelet coefficients as features for the epoch.

25. The method for diagnosing the presence of sleep disorders of claim 8, selecting a subset of features for use by the classifier.

26. The method for diagnosing the presence of sleep disorders of claim 25, employing an algorithm to select the best subset based on the training data.

27. The method for diagnosing the presence of sleep disorders of claim 26, wherein the algorithm is selected from the group consisting of best first, remove worst, random start with add and remove, simulated annealing and genetic algorithms.

28. The method for diagnosing the presence of sleep disorders of claim 26, employing cross-validation to select the best subset comprising

splitting the training data into groups or folds,

performing multiple accuracy tests,

using in each case a subset of the folds for training,

testing the resulting classifier for accuracy on the remaining fold(s).

29. The method for diagnosing the presence of sleep disorders of claim 25, training the classifier using the training data set.

30. The method for diagnosing the presence of sleep disorders of claim 29, wherein the classifier is selected from the group consisting of Baysean maximum likelihood linear and quadratic discriminants, neural networks and support vector machines.

31. The method for diagnosing the presence of sleep disorders of claim 30, further comprising calculating a discriminant function,

providing probability estimates for different classes by applying the discriminant function to features calculated from new data to be classified, and

assigning the data (epoch) to the class with the highest probability.

32. The method for diagnosing the presence of sleep disorders of claim 31, wherein the discriminant function comprises two weight vectors (of the same length as the feature subset) and two constants.

33. The method for diagnosing the presence of sleep disorders of claim 31, wherein the discriminant functions and probability are calculated as follows:

setting the probability cutoff to suit the desired sensitivity and specificity.

34. (canceled)

35. (canceled)

36. (canceled)

37. (canceled)

38. (canceled)

39. (canceled)

40. (canceled)

41. (canceled)

42. (canceled)

43. (canceled)

44. (canceled)

45. (canceled)

46. (canceled)

47. (canceled)

48. (canceled)

49. (canceled)

50. (canceled)

51. (canceled)