Method for detecting bio signal features in the presence of noise

• US 8,073,218 B2
• Filed: 09/25/2008
• Issued: 12/06/2011
• Est. Priority Date: 09/25/2008
• Status: Expired due to Fees

• Current Assignee: Air Products and Chemicals Incorporated (Linde Plc)
• Original Assignee: Air Products and Chemicals Incorporated (Linde Plc)
• Inventors: Toth, Michael S., Petit, Cecilia Anna Paulette
• Primary Examiner(s): Tang, Minh N
• Application Number:
  US12/284,999
• Publication Number:
  US 20100074482A1
• Time in Patent Office: 1,167 Days
• Field of Search: None
• US Class Current: 382/128
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/7203 for noise prevention, reduction or removal
• CPC Class Codes: A61B 5/7239 using differentiation including higher order derivatives
• CPC Class Codes: A61B 5/726 using Wavelet transforms
• CPC Class Codes: G06F 2218/06 by applying a scale-space analysis, e.g. using wavelet analysis

Abstract

A computer-based method for detecting features in a digitally sampled biological (bio) signal includes the steps of applying a wavelet transform to the bio signal to generate a list of wavelet transform coefficients, ranking the coefficients according to value or frequency, and removing at least one low value or low frequency coefficient. The method further includes the steps of performing an inverse wavelet transform using the remaining coefficients to generate a reconstructed signal, and detecting feature locations using the reconstructed signal.

Claims

1. A computer-based method for detecting features in a digitally sampled biological (bio) signal, comprising:

applying a wavelet transform to the bio signal to generate a list of wavelet transform coefficients;

ranking the coefficients according to value;

removing at least one low value coefficient;

performing an inverse wavelet transform using the remaining coefficients to generate a reconstructed signal; and

detecting feature locations using the reconstructed signal.

2. The method of claim 1, further comprising:

applying a length transform; and

detecting feature locations from an output of the length transform.

3. The method of claim 2, wherein the length transform is applied to the reconstructed signal.

4. The method of claim 2, wherein the length transform is applied to a mathematical derivative of the reconstructed signal.

5. The method of claim 1, wherein the feature locations are detected by locating a peak maximum.

6. The method of claim 2, wherein the feature locations are detected by:

determining a maximum value of the length transform output;

locating a first point in a left portion of a peak pulse of the length transform output and a second point in a right portion of the peak pulse, the first and second points being equal in value, the value being a function of the maximum value; and

detecting feature locations from the length transform output by calculating a location halfway between the first and second points.

7. The method of claim 6, wherein the function corresponds to half of the maximum value.

8. The method of claim 2, further comprising:

rectifying an input of the length transform.

9. The method of claim 2, further comprising:

selecting a window size of the length transform as a function of a known feature duration.

10. The method of claim 1, further comprising:

prior to the step of detecting, removing baseline drift from the bio signal.

11. The method of claim 1, further comprising:

initially removing all but a set of highest value coefficients;

determining whether a result of a current detection is satisfactory; and

if the result is unsatisfactory, repeating the steps of adding a next highest coefficient to the remaining coefficients, performing the inverse wavelet transform, and detecting until the result is satisfactory.

12. The method of claim 11, wherein the determination of satisfaction is based on whether a number of features detected is within a range of expected values.

13. The method of claim 11, wherein the determination of satisfaction is based on whether a change occurs in a detected feature rate over time.

14. The method of claim 1, further comprising:

initially removing all but a set of highest value coefficients;

adding a next highest coefficient to the remaining coefficients;

repeating the steps of performing the inverse wavelet transform and detecting;

determining whether a result of a current detection is a substantial improvement compared to a result of a previous detection; and

if the current result is a substantial improvement, repeating the steps of adding a next highest coefficient, performing the inverse wavelet transform, and detecting until a result of the current detection no longer yields a substantial improvement compared to a previous detection.

15. The method of claim 14, wherein the determination of improvement is based on a rate of change in a number of features detected when an additional coefficient is added.

16. The method of claim 1, wherein the bio signal is an electrocardiogram signal and the features are heartbeats.

17. A computer-based method for detecting features in a digitally sampled biological (bio) signal, comprising:

applying a wavelet transform to the bio signal to generate a list of wavelet transform coefficients;

ranking the coefficients according to frequency;

removing at least one low frequency coefficient;

performing an inverse wavelet transform using the remaining coefficients to generate a reconstructed signal; and

detecting feature locations using the reconstructed signal.

18. The method of claim 17, wherein the feature locations are detected by locating a peak maximum.

19. The method of claim 17, further comprising:

filtering the reconstructed signal through a low pass filter; and

detecting feature locations from the filtered signal.

20. The method of claim 19, wherein the feature locations are detected by:

determining a maximum value of the filtered signal;

locating a first point in a left portion of a peak pulse of the filtered signal and a second point in a right portion of the peak pulse, the first and second points being equal in value, the value being a function of the maximum value; and

detecting feature locations from the filtered signal by calculating a location halfway between the first and second points.

21. The method of claim 20, wherein the function corresponds to half of the maximum value.

22. The method of claim 19, further comprising:

rectifying the reconstructed signal prior to filtering.

23. The method of claim 19, further comprising:

rectifying the filtered signal after filtering.

24. The method of claim 17, further comprising:

initially removing all but a set of highest frequency coefficients;

determining whether a result of a current detection is satisfactory; and

if the result is unsatisfactory, repeating the steps of adding a next highest coefficient to the remaining coefficients, performing the inverse wavelet transform, and detecting feature locations until the result is satisfactory.

25. The method of claim 24, wherein the determination of satisfaction is based on whether a number of features detected is within a range of expected values.

26. The method of claim 24, wherein the determination of satisfaction is based on whether a change occurs in a detected feature rate over time.

27. The method of claim 17, further comprising:

initially removing all but a set of highest frequency coefficients;

adding a next highest coefficient to the remaining coefficients;

repeating the step of detecting feature locations;

determining whether a result of a current detection is a substantial improvement compared to a result of a previous detection; and

if the current result is a substantial improvement, repeating the steps of adding a next highest coefficient, performing an inverse wavelet transform, and detecting until a result of the current detection no longer yields a substantial improvement compared to a previous detection.

28. The method of claim 27, wherein the determination of improvement is based on a rate of change in a number of features detected when an additional coefficient is added.

29. The method of claim 17, wherein the bio signal is an electrocardiogram signal and the features are heartbeats.

30. A device for detecting features in a digitally sampled biological (bio) signal, comprising:

a communications arrangement configured to receive the bio signal;

a processor; and

a memory including instructions configuring the processor to;

apply a wavelet transform to the bio signal to generate a list of wavelet transform coefficients;

rank the coefficients according to value;

remove at least one low value coefficient;

perform an inverse wavelet transform using the remaining coefficients to generate a reconstructed signal; and

detect feature locations using the reconstructed signal.

31. A device for detecting features in a digitally sampled biological (bio) signal, comprising:

a communications arrangement configured to receive the bio signal;

a processor; and

a memory including instructions configuring the processor to;

apply a wavelet transform to the bio signal to generate a list of wavelet transform coefficients;

rank the coefficients according to frequency;

remove at least one low frequency coefficient;

perform an inverse wavelet transform using the remaining coefficients to generate a reconstructed signal; and

detect feature locations using the reconstructed signal.

32. A computer-readable medium having stored thereon a series of instructions executable by a processor for detecting features in a digitally sampled biological (bio) signal, the instructions configured to cause the processor to perform the steps of:

applying a wavelet transform to the bio signal to generate a list of wavelet transform coefficients;

ranking the coefficients according to value;

removing at least one low value coefficient;

performing an inverse wavelet transform using the remaining coefficients to generate a reconstructed signal; and

detecting feature locations using the reconstructed signal.

33. A computer-readable medium having stored thereon a series of instructions executable by a processor for detecting features in a digitally sampled biological (bio) signal, the instructions configured to cause the processor to perform the steps of:

applying a wavelet transform to the bio signal to generate a list of wavelet transform coefficients;

ranking the coefficients according to frequency;

removing at least one low frequency coefficient;

performing an inverse wavelet transform using the remaining coefficients to generate a reconstructed signal; and

detecting feature locations using the reconstructed signal.