Separation of a subcutaneous cardiac signal from a plurality of composite signals
A cardiac monitoring and/or stimulation method and systems provide monitoring, defibrillation and/or pacing therapies, including systems detecting and/or treating cardiac arrhythmia. A system includes a housing coupled to a plurality of electrodes configured for subcutaneous non-intrathoracic sensing. A signal processor receives a plurality of composite signals associated with a plurality of sources, separates a signal from the plurality of composite signals using blind source separation, and identifies a cardiac signal. The signal processor may iteratively separate signals from the plurality of composite signals until the cardiac signal is identified. A method of signal separation includes detecting a plurality of composite signals at a plurality of locations, separating a signal using blind source separation, and identifying a cardiac signal. The separation may include a principal component analysis and/or an independent component analysis. The composite signals may be filtered before separation using band-pass filtering, adaptive, or other filters.
1. A signal separation method, comprising:
detecting a plurality of composite signals at a plurality of subcutaneous non-intrathoracic locations, the plurality of composite signals associated with a plurality of sources;
separating a signal from the plurality of composite signals using blind source separation; and
identifying the separated signal as a cardiac signal.
2. The method of claim 1 , further comprising iteratively separating signals from the plurality of composite signals until the cardiac signal is identified.
3. The method of claim 1 , wherein the plurality of composite signals comprises the cardiac signal and at least one non-cardiac signal.
4. The method of claim 1 , wherein the blind source separation comprises a principal component analysis.
5. The method of claim 4 , wherein the principal component analysis includes a singular value decomposition.
6. The method of claim 4 , wherein the principal component analysis includes an eigenvalue decomposition.
7. The method of claim 1 , wherein the blind source separation comprises an independent component analysis.
8. The method of claim 1 , wherein the blind source separation comprises a principal component analysis and an independent component analysis.
9. The method of claim 1 , further comprising filtering each of the plurality of composite signals before separating the signal.
10. The method of claim 1 , further comprising band-pass filtering each of the plurality of composite signals.
11. The method of claim 1 , further comprising determining a cardiac condition using the cardiac signal.
12. The method of claim 1 , further comprising detecting a cardiac condition using the separated signal, wherein detecting the cardiac condition comprises performing a rate based analysis of the cardiac signal.
13. The method of claim 1 , further comprising detecting a cardiac condition using the separated signal, wherein detecting the cardiac condition comprises performing a morphology based analysis of the cardiac signal.
14. The method of claim 1 , further comprising detecting a cardiac condition using the separated signal, wherein detecting the cardiac condition comprises performing a pattern and rate based analysis of the cardiac signal.
15. The method of claim 1 , further comprising detecting a cardiac condition using the cardiac signal and implantably treating the cardiac condition.
16. The method of claim 15 , wherein implantably treating the cardiac condition comprises delivering a cardiac stimulation therapy.
17. The method of claim 15 , wherein implantably treating the cardiac condition comprises delivering a cardiac stimulation therapy using an implanted subcutaneous non-intrathoracic electrode.
18. The method of claim 15 , wherein detecting the plurality of composite signals at the plurality of subcutaneous locations comprises using a subcutaneous non-intrathoracic electrode array and wherein implantably treating the cardiac condition comprises using the electrode array to provide a cardiac stimulation therapy.
19. The method of claim 1 , further comprising:
forming a composite signal matrix from the plurality of composite signals; and
performing a principal component analysis on the composite signal matrix, thereby producing a set of eigenvalues and associated eigenvectors.
20. The method of claim 19 , wherein an eigenvector of the associated eigenvectors that is associated with a largest magnitude eigenvalue of the set of eigenvalues is used to separate the signal from the plurality of composite signals.
21. The method of claim 20 , wherein the signal is separated from the plurality of composite signals by multiplying the composite signal matrix by the eigenvector.
22. The method of claim 20 , wherein the signal is separated iteratively from the plurality of composite signals using an eigenvector associated with a next-largest magnitude eigenvalue for each iteration, until the cardiac signal is identified.
23. The method of claim 22 , wherein the cardiac signal is identified using a local peak density of the separated signal, wherein the local peak density is within a predetermined range.
24. The method of claim 22 , wherein the cardiac signal is identified using beat detection on the separated signal, wherein a beat rate is within a predetermined range.
25. The method of claim 22 , wherein the cardiac signal is identified using a local rate of occurrence of significant points in the separated signal, wherein the local rate of occurrence is within a predetermined range.
26. The method of claim 22 , wherein the cardiac signal is identified using a morphology of the separated signal, wherein the morphology satisfies a physiological characteristic.
27. The method of claim 19 , further comprising forming an intermediate signal matrix by multiplying the composite signal matrix by a matrix composed of the eigenvectors corresponding to at least the two largest eigenvalues, and finding a unitary matrix corresponding to a rotational transformation of the intermediate signal matrix that increases the independence of the intermediate signals.
28. The method of claim 27 , further comprising forming an independent signal matrix by multiplying the intermediate signal matrix by the unitary matrix, wherein the independent signal matrix comprises a set of independent signal vectors.
29. The method of claim 28 , wherein a signal vector from the set of independent signal vectors is iteratively examined until a cardiac signal is identified.
30. The method of claim 29 , wherein the cardiac signal is identified using a local peak density of the examined signal, wherein the local peak density is within a predetermined range.
31. The method of claim 29 , wherein the cardiac signal is identified using beat detection on the examined signal, wherein a beat rate is within a predetermined range.
32. The method of claim 29 , wherein the cardiac signal is identified using a local rate of occurrence of significant points in the examined signal, wherein the local rate of occurrence is within a predetermined range.
33. The method of claim 29 , wherein the cardiac signal is identified using a morphology of the examined signal, wherein the morphology satisfies a physiological characteristic.
34. The method of claim 1 , further comprising forming an estimate of a spatial covariance matrix for a composite signal matrix formed from the plurality of composite signals.
35. The method of claim 34 , further comprising performing a principal component analysis on the estimate of the spatial covariance matrix, producing a set of eigenvalues and associated eigenvectors.
36. The method of claim 35 , wherein an eigenvector of the associated eigenvectors that is associated with a largest magnitude eigenvalue of the set of eigenvalues is used to separate the signal from the plurality of composite signals.
37. The method of claim 36 , wherein the signal is separated from the plurality of composite signals by multiplying the composite signal matrix by the eigenvector.
38. The method of claim 36 , wherein the signal is separated iteratively from the plurality of composite signals using an eigenvector associated with a next-largest magnitude eigenvalue for each iteration, until the cardiac signal is identified.
39. The method of claim 38 , wherein the cardiac signal is identified using a local peak density of the separated signal, wherein the local peak density is within a predetermined range.
40. The method of claim 38 , wherein the cardiac signal is identified using beat detection on the separated signal, wherein a beat rate is within a predetermined range.
41. The method of claim 38 , wherein the cardiac signal is identified using a local rate of occurrence of significant points in the separated signal, wherein the local rate of occurrence is within a predetermined range.
42. The method of claim 38 , wherein the cardiac signal is identified using a morphology of the separated signal, wherein the morphology satisfies a physiological characteristic.
43. The method of claim 35 , further comprising forming an intermediate signal matrix by multiplying the composite signal matrix by a matrix composed of the eigenvectors corresponding to at least the two largest eigenvalues, and finding a unitary matrix corresponding to a rotational transformation of the intermediate signal matrix that increases the independence of the intermediate signals.
44. The method of claim 43 , further comprising forming an independent signal matrix by multiplying the intermediate signal matrix by the unitary matrix, wherein the independent signal matrix comprises a set of independent signal vectors.
45. The method of claim 44 , wherein a signal vector from the set of independent signal vectors is iteratively examined until a cardiac signal is identified.
46. The method of claim 45 , wherein the cardiac signal is identified using a local peak density of the examined signal, wherein the local peak density is within a predetermined range.
47. The method of claim 45 , wherein the cardiac signal is identified using beat detection on the examined signal, wherein a beat rate is within a predetermined range.
48. An implantable cardiac device, comprising:
a plurality of electrodes configured for subcutaneous non-intrathoracic sensing; and
a signal processor coupled to the plurality of electrodes, the signal processor configured to receive a composite signal associated with a plurality of sources sensed by at least two of the electrodes, and to extract a cardiac signal from the plurality of composite signals using blind source separation.
49. The device of claim 48 , further comprising at least one lead coupled to the plurality of subcutaneous non-intrathoracic electrodes.
50. The device of claim 48 , further comprising a plurality of leads coupled to the plurality of subcutaneous non-intrathoracic electrodes.
51. The device of claim 48 , further comprising a plurality of leads, wherein at least one of the plurality of leads is coupled to an array of electrodes.
52. The device of claim 48 , wherein the signal processor further comprises an adaptive filter, the adaptive filter altering its filtering characteristics based on a level of ambient noise.
53. The device of claim 52 , wherein the level of ambient noise is determined using the independent component analysis.
54. The device of claim 48 , wherein the blind source separation includes a principal component analysis that produces a set of eigenvalues and associated eigenvectors, and an eigenvector with a largest magnitude eigenvalue is used to extract the cardiac signal from the plurality of composite signals.
55. The device of claim 54 , wherein a signal is separated iteratively from the plurality of composite signals using a next-largest magnitude eigenvalue for each iteration, until the cardiac signal is extracted from the plurality of composite signals.
56. An implantable cardiac device, comprising:
means for subcutaneously detecting a plurality of composite signals produced by a plurality of sources;
means for separating a signal from the plurality of composite signals “using blind source separation”;and
means for identifying the separated signal as a cardiac signal.
57. The device of claim 56 , wherein the separating means performs a principal component analysis on the plurality of composite signals.
58. The device of claim 56 , wherein the separating means produces a set of eigenvalues and associated eigenvectors, and an eigenvector with a largest magnitude eigenvalue is used for separating the signal from the plurality of composite signals.
59. The device of claim 58 , wherein the separating means iteratively separates the signal from the plurality of composite signals using a next-largest magnitude eigenvalue for each iteration, until the cardiac signal is separated from the plurality of composite signals.
60. The device of claim 56 , further comprising means for detecting a cardiac condition using the cardiac signal.
61. The device of claim 60 , further comprising means for treating the cardiac condition.