Method and system for the denoising of large-amplitude artifacts in electrograms using time-frequency transforms

• US 7,672,717 B1
• Filed: 10/19/2004
• Issued: 03/02/2010
• Est. Priority Date: 10/22/2003
• Status: Active Grant

• Current Assignee: Bionova Technologies Incorporated
• Original Assignee: Bionova Technologies Incorporated
• Inventors: Bibian, Stephane, Huzmezan, Mihai, Zikov, Tatjana, Dumont, Guy Albert, Ries, Craig Robert
• Primary Examiner(s): Marmor, II; Charles A
• Assistant Examiner(s): Natnithithadha; Navin
• Application Number:
  US10/968,348
• Time in Patent Office: 1,960 Days
• Field of Search: 600/300, 600/544, 600/545, 600508-526, 600/546, 600/547
• US Class Current: 600/544
• CPC Class Codes: A61B 5/369 Electroencephalography [EEG] devices for psychotechnics A61B5/16
• CPC Class Codes: A61B 5/398 Electrooculography [EOG], e.g. detecting nystagmus; Electroretinography [ERG]
• CPC Class Codes: A61B 5/726 using Wavelet transforms

Abstract

The present invention relates to a method of signal processing of electrograms for use in medical devices, preferably by time-frequency transforms. The present invention additionally relates to a system for receiving and analyzing such signals. The present invention preferably is a method utilizing the time-frequency transforms, such as wavelet transforms, for the purpose of artifact removal from EGs. These transforms decompose a signal in both time and frequency domains, and therefore, are well suited for non-stationary signal analysis. As a result, dissimilar signal features are well localized both in time and frequency, which potentially provides a good separation between the signal of interest and artifacts. This particularly applies to large-amplitude artifacts corrupting EGs.

Claims

1. A system for removing artifacts from an observed electrogram of a subject in order to produce a denoised electrogram, given a time-frequency transform, and given a database of reference electrograms, each said reference electrogram being free of said artifacts, and each said reference electrogram having been acquired from individuals other than said subject, said system comprising:

a. a sensor for acquiring the electrogram of said subject to complement said database of reference electrograms, and for producing said observed electrogram; and

b. a digital signal processor for;

i. applying said time-frequency transform to said reference electrograms to produce a plurality of reference coefficient sets, each said reference coefficient set corresponding to a respective time-frequency span;

ii. calculating a threshold for each said time-frequency span based on a corresponding reference coefficient set;

iii. applying said time-frequency transform to said observed electrogram to produce a plurality of observed coefficient sets, each said observed coefficient set corresponding to a respective time-frequency span;

iv. comparing each coefficient of each said observed coefficient set with the corresponding threshold for a respective time-frequency span;

v. replacing every large coefficient of each said observed coefficient set which is above the respective threshold for the corresponding time-frequency span with a replacement value in order to produce denoised observed coefficient sets; and

vi. applying the inverse transform of said time-frequency transform to said denoised observed coefficient sets to reconstruct the denoised electrogram.

2. The system of claim 1, wherein said electrograms are electroencephalograms.

3. The system of claim 1, wherein said time-frequency transform is a wavelet transform.

4. The system of claim 1, wherein said threshold comprises two values which define a maximum coefficient bound and a minimum coefficient bound, and wherein said maximum and minimum coefficient bounds are defined such that most of coefficients of said reference coefficient sets are contained within these bounds.

5. The method of claim 1, wherein said replacement value is any number less than or equal to the threshold including 0.

6. The method of claim 1, wherein said replacement value is an estimation of the value of the coefficient being replaced, had the artifact never occurred.

7. A method for removing artifacts from an observed electrogram of a subject in order to provide a denoised electrogram, said method comprising:

a. acquiring a plurality of reference electrograms, each said reference electrogram being free of said artifacts;

b. selecting a time-frequency transform which, when applied to said reference electrograms yields a plurality of reference coefficient sets, each said reference coefficient set corresponding to a respective time-frequency span;

c. selecting a threshold for each time-frequency span;

d. acquiring an electrogram from said subject to produce said observed electrogram;

e. applying said time-frequency transform to said observed electrogram to produce a plurality of observed coefficient sets, each said observed coefficient set corresponding to a respective time-frequency span;

f. comparing each coefficient of each said observed coefficient set with the corresponding threshold for a respective time-frequency span; and

g. replacing every large coefficient of each said observed coefficient set which is above exceeds the respective threshold for the corresponding time-frequency span with a replacement value in order to produce denoised observed coefficient sets.

8. The method of claim 7, wherein said electrograms are electroencephalograms.

9. The method of claim 7, wherein said time-frequency transform is a wavelet transform.

10. The method of claim 7, wherein said threshold comprises two values which define a maximum coefficient bound and a minimum coefficient bound, and wherein said maximum and minimum coefficient bounds are defined such that most of coefficients of said reference coefficient sets are contained within these bounds.

11. The method of claim 7, wherein said replacement value is any number less than or equal to the threshold, including 0.

12. The method of claim 7, wherein said replacement value is an estimation of the value of the coefficient being replaced, had the artifact never occurred.

13. The method of claim 7 wherein the inverse transform of said selected time-frequency transform is applied to said denoised observed coefficient sets to reconstruct a denoised electrogram.

14. The method of claim 7 wherein said time-frequency transform is a redundant wavelet transform.

15. The method of claim 7 wherein said time-frequency transform is a redundant wavelet packet transform.

16. The method of claim 7 wherein said reference electrograms are acquired from at least one individual other than said subject.

17. A method for extracting artifacts from an observed electrogram of a subject in order to provide an artifactual electrogram, said method comprising:

a. acquiring a plurality of reference electrograms, each said reference electrogram being free of said artifacts;

b. selecting a time-frequency transform which, when applied to said reference electrograms yields a plurality of reference coefficient sets, each said reference coefficient set corresponding to a respective time-frequency span;

c. selecting a threshold for each time-frequency span;

d. acquiring an electrogram from said subject to produce said observed electrogram;

e. applying said time-frequency transform to said observed electrogram to produce a plurality of observed coefficient sets, each said observed coefficient set corresponding to a respective time-frequency span;

f. comparing each coefficient of each said observed coefficient set with the corresponding threshold for a respective time-frequency span;

g. replacing every small coefficient of each said observed coefficient set which is less than the respective threshold for the corresponding time-frequency span with a replacement value in order to produce artifactual observed coefficient sets; and

h. applying the inverse transform of said selected time-frequency transform to said artifactual observed coefficient sets to reconstruct the artifactual electrogram.

18. The method of claim 17, wherein said electrograms are electroencephalograms.

19. The method of claim 17, wherein said time-frequency transform is a wavelet transform.

20. The method of claim 17, wherein said threshold comprises two values which define a maximum coefficient bound and a minimum coefficient bound, and wherein said maximum and minimum coefficient bounds are defined such that most of coefficients of said reference coefficient sets are contained within these bounds.

21. The method of claim 17, wherein said replacement value is 0.

22. The method of claim 17 wherein said reference electrograms are acquired from at least one individual other than said subject.

23. The method of claim 17 wherein said artifactual electrogram is subtracted from said observed electrogram in order to yield a denoised electrogram.