Patient signal filtering

1. A non-transitory computer readable medium having embodied thereon computer-executable instructions which, when executed, cause a computer to perform steps for signal filtering, the steps comprising:

receiving an electrocardiogram signal, wherein the electrocardiogram signal includes an electrocardiogram signal component, noise, a respiration signal component and a blood pressure signal component;

removing the electrocardiogram signal component from the electrocardiogram signal to generate a first output signal by performing an electrocardiogram signature cycle matching pursuit method;

eliminating the noise from the first output signal to generate a second output signal by performing sub-bandwidth filtering of the first output signal;

removing the respiration signal component from the second output signal to generate a third output signal; and

reconstructing a time-domain blood pressure signal based on the third output signal by performing inverse fast Fourier transform (IFFT) on integrated frequency components of the third output signal.

2. The non-transitory computer readable medium of claim 1 having embodied thereon further computer-executable instructions which, when executed, cause the computer to perform the electrocardiogram signature cycle matching pursuit method by deriving a signature cycle from the electrocardiogram signal and removing the electrocardiogram signal component associated with the signature cycle from the electrocardiogram signal.

3. The non-transitory computer readable medium of claim 2 having embodied thereon further computer-executable instructions which, when executed, cause the computer to perform the sub-bandwidth filtering of the first output signal by performing steps comprising:

separating, via multiple filters, the first output signal into multiple frequency sub-bandwidths to generate the second output signal; and

adaptively adjusting, in response to changes in the second output signal, filter parameters associated with the filters to remove the noise.

4. The non-transitory computer readable medium of claim 1 having embodied thereon further computer-executable instructions which, when executed, cause the computer to remove the respiration signal component by low pass filtering.

5. The non-transitory computer readable medium of claim 1 having embodied thereon further computer-executable instructions which, when executed, cause the computer to clean up the blood pressure signal by applying a Savitzky-Golay smoothing filter.

6. A method of signal filtering, comprising:

receiving an electrocardiogram signal that includes first, second, third signal components and a signal component of interest;

removing the first signal component from the electrocardiogram signal to generate a first output signal by performing a signature cycle matching pursuit method based on the electrocardiogram signal, wherein performing the signature cycle matching pursuit method includes

selecting, from the electrocardiogram signal, reference cycles,

averaging the reference cycles to generate averaged signal cycles,

deriving a signature cycle from the averaged signal cycles, and

removing the first signal component associated with the signature cycle from the electrocardiogram signal;

removing the second signal component from the first output signal to generate a second output signal by performing linear sub-bandwidth filtering of the first output signal;

removing the third signal component from the second output signal to generate a third output signal; and

reconstructing a signal of interest in time domain based on the third output signal by performing inverse fast Fourier transform (IFFT) on integrated frequency components of the third output signal.

7. The method of claim 6 wherein the electrocardiogram signal comprises a blood pressure signal component, an electrocardiogram signal component, a respiration signal component and a noise component.

8. The method of claim 7 wherein the signal of interest is a blood pressure signal component.

9. The method of claim 6 wherein removing the first signal component associated with the signature cycle from the electrocardiogram signal comprises synchronizing the signature cycle with an R wave of at least one cycle of the electrocardiogram signal.

10. The method of claim 6 wherein the signature cycle comprises an electrocardiogram signature cycle.

11. The method of claim 6 wherein deriving the signature cycle comprises:

reconstructing the signature cycle by performing a wavelet-based time-frequency distribution analysis.

12. The method of claim 11 wherein performing the wavelet-based time-frequency distribution analysis comprises:

segmenting a time-frequency distribution of the signal component into a set of sub-components;

selecting, based on time duration and frequency range determination, a sub-set of six clinically significant sub-components from the set of sub-components; and

reconstructing the signature cycle based on the sub-set of the six clinically significant sub-components.

13. The method of claim 6 wherein performing the sub-bandwidth filtering of the first output signal comprises:

separating, via multiple filters, the first output signal into multiple frequency sub-bandwidths to generate the second output signal; and

adaptively adjusting, in response to changes in the second output signal, filter parameters associated with the filters to remove the second signal component.

14. The method of claim 13 wherein the filter parameters comprise a number of filters, bandwidth size, order, shape, amplification ratio, cut-off frequency, or a combination thereof.

15. The method of claim 13 wherein adaptively adjusting the filter parameters comprises performing an artificial neural network-based method.

16. The method of claim 6 wherein removing the third signal component from the second output signal comprises removing a respiration signal component by performing low pass filtering.

17. The method of claim 6 further comprises cleaning up the signal of interest by applying a Savitzky-Golay smoothing filter.