Spectrum analysis of coronary artery turbulent blood flow

• US 20100036254A1
• Filed: 08/09/2008
• Published: 02/11/2010
• Est. Priority Date: 08/09/2008
• Status: Active Grant

• Current Assignee: Phonoflow Medical, LLC
• Original Assignee: Phonoflow Medical, LLC
• Inventors: Zaorski, Ralph Walter, Norris, Roger Paul, Owsley, Norman Lee
• Granted Patent: US 8,419,651 B2
• Time in Patent Office: Days
• US Class Current: 600/454
• CPC Class Codes: A61B 2562/0204 Acoustic sensors
• CPC Class Codes: A61B 5/026 Measuring blood flow A61B3/1233, A61B3/1241 take precedence
• CPC Class Codes: A61B 5/7257 using Fourier transforms

Abstract

A method includes acquiring vibrational cardiac data from an array of N transducers wherein the transducers are coupled to a human. A master replica is selected from a segment of the vibrational cardiac data. The master replica is correlated with the segment to obtain a plurality of local maxima. Vibrational cardiac data that were emitted during a diastolic interval are extracted from each heart cycle with the aid of the plurality of local maxima. A two-dimensional space-time frequency power spectrum is processed for Equivalent Rectangular Bandwidth, which provides estimates of the energy produced by turbulent blood flow through a coronary stenosis.

Claims

1. A method comprising:

acquiring vibrational cardiac data from a transducer wherein the transducer is coupled to a human;

selecting a master replica from a segment of the vibrational cardiac data;

correlating the master replica with the segment to obtain a plurality of local maxima; and

extracting vibrational cardiac data that were emitted during a diastolic interval from each heart cycle with the aid of the plurality of local maxima.

2. The method of claim 1, further comprising:

filtering the segment before the selecting, wherein the filtering performs a band-pass filter operation on the segment.

3. The method of claim 2, further comprising:

performing an envelope detection on the segment before the filtering.

4. The method of claim 1, further comprising:

computing an average amplitude for the vibrational cardiac data obtained during the extracting; and

discarding vibrational cardiac data that deviates by more than a specified amount from the average amplitude.

5. The method of claim 1, wherein the master replica is a portion of the segment.

6. The method of claim 5, wherein the portion is selected from the group consisting of a fraction of a heart cycle, one heart cycle, and more than one heart cycle.

7. The method of claim 1, further comprising:

picking a high quality transducer channel from an array of N transducer channels, wherein the vibrational cardiac data is collected from the high quality transducer channel.

8. The method of claim 7, wherein a number of heart cycles collected is approximately equal to ten to fifteen times N.

9. The method of claim 1, wherein the selecting is performed by a human operator.

10. The method of claim 1, wherein vibrational cardiac data contains a plurality of segments.

11. The method of claim 10, wherein the plurality is equal to six segments.

12. The method of claim 10, wherein the human starts the acquiring to coincide with breath-hold and stops the acquiring to coincide with non-breath hold.

13. The method of claim 1, wherein the segment includes a continuous period of time during which the human can comfortably refrain from breathing.

14. The method of claim 1, wherein the segment contains approximately 120 heart cycles.

15. The method of claim 1, wherein the segment contains approximately 20 heart cycles.

16. The method of claim 1, wherein the extracting further comprises:

discarding a local maxima whose amplitude is below an amplitude threshold value.

17. The method of claim 16, wherein the extracting further comprises:

comparing a separation in time between adjacent local maxima against a time threshold value and discarding the latter local maxima when the separation is below the time threshold value.

18. The method of claim 1, wherein the extracting is assisted by a human operator and the human operator sets a start time and optionally an end time for the diastolic interval.

19. The method of claim 1, further comprising:

computing a mean power level for all vibrational cardiac data collected during diastolic intervals in a segment;

comparing an average power level for vibrational cardiac data collected during a diastolic interval to the mean power level and discarding the vibrational cardiac data if the average power level deviates from the mean power level by a predetermined amount.

20. A method for processing acquired synchronized diastolic interval vibrational cardiac data (DI data) collected by an array of transducers during a plurality of heart cycles from a human, comprising:

performing a time-to-frequency transformation on the DI data, wherein the performing is applied to the DI data acquired from each transducer during the plurality of heart cycles to obtain a plurality of transducer channel frequency complex Fourier spectra;

calculating a cross-channel spectral density matrix (CSDM) from the plurality of transducer frequency spectra, wherein the CSDM is calculated as a time average over the time synchronized plurality of heart cycles;

performing an eigenvalue-eigenvector decomposition (EED) of the CSDM to obtain a spatial frequency number versus temporal frequency representation of the DI data.

21. The method of claim 20, further comprising:

partitioning a diastolic interval into S time slots, wherein the performing the time-to-frequency transformation is done separately within each slot number of the S time slots.

22. The method of claim 21, wherein S is selected from the group consisting of 1, 2, 3, 4, 5, and 6.

23. The method of claim 22, wherein the duration of a time slot is approximately one-third the duration of a diastolic interval and adjacent time slots may overlap in time.

24. The method of claim 20, wherein the time-to-frequency transformation is a Fast Fourier Transform (FFT) operation.

25. The method of claim 22, wherein a time-to-frequency transformation is performed for the slot number within each heart cycle of the plurality of time synchronized heart cycles, to produce a plurality of time-to-frequency transformations, then the plurality of time-to-frequency transformations are averaged across the plurality of heart cycles for the slot number.

26. The method of claim 20, wherein the time average extends over more than one segment.

27. The method of claim 26, wherein a new master replica has been selected from each segment in order to obtain the DI data.

28. The method of claim 21, wherein the time-to-frequency transformation is a discrete Fast Fourier Transform (DFFT) operation and the CSDM (R(k, m)) is calculated as a time average over a heart beat count index b, and separately for each frequency bin “

k” and

slot number “

m,”

with N equal to the number of vibration transducer channels, according to;

29. The method of claim 28, wherein the EED of the DI data is calculated as:

[M(k, m), L(k, m), M(k, m)]=svd(R(k, m));

where

M(k, m) is the N-by-N matrix of orthonormal eigenvectors of R(k, m) as columns and L(k, m) is the diagonal matrix of corresponding eigenvalues in monotonic decreasing order and “

svd”

indicates “

Singular Value Decomposition.”

30. The method of claim 29, further comprising:

computing a two dimensional Temporal and Spatial Spectrum (TSS) noise floor λ

₀ of the DI data, wherein DI data for a subset N_f of eigenvalues of R(k,m) are averaged over all FFT frequency values to produce a subset of frequency averages, then the subset of frequency averages are averaged over the subset N_f of eigenvalues, thereby producing the two dimensional Temporal and Spatial Spectrum (TSS) noise floor λ

₀ of the DI data.

31. The method of claim 30, wherein a size of the subset N_f of eigenvalues is between 10 and 30 percent of N.

32. The method of claim 30, further comprising:

performing a summation λ

_s(p), over frequency, of eigenvalues as a function of frequency, for a fixed value of p, on the DI data, for the C_s(p) eigenvalues of the DI data that exceed a threshold given by α

λ

_0,; and

the summation proceeds from p=1, 2, . . . , N-N_f; and

performing a summation λ

_n(p), over frequency, of eigenvalues as a function of frequency, for a fixed value of p, on the DI data, for the eigenvalues of the DI data that do not exceed the threshold given by α

λ

₀.

33. The method of claim 20, further comprising:

correcting the DI data for artifacts of 60 cycle/sec and harmonic, room ambient electromagnetic interference before the calculating by nulling and extrapolating the frequency spectrum through a 60 cycle artifact in at least one of the plurality of transducer frequency complex Fourier spectra.

34. The method of claim 32 further comprising:

computing an Equivalent Rectangular Bandwidth (ERB) of the DI data for the p-th spatial eigenvalue, wherein the ERB=C_s(p)*(FFT frequency resolution), and C_s(p) is a counter used during the summation λ

_s(p).

35. The method of claim 29, further comprising:

presmoothing a frequency spectrum of the EED; and

calculating an Equivalent Rectangular Bandwidth (ERB) of the DI data by identifying the frequency at which the LLS fit decreases to a specified level with increasing frequency.

36. An apparatus comprising:

a data processing device configured to accept acquired diastolic interval vibrational cardiac data (DI data) collected by an array of transducers during a plurality of heart cycles; and

a computer readable medium containing executable computer program instructions, which when executed by the data processing system, cause the data processing system to perform a method comprising;

partitioning a diastolic interval into S time slots;

performing a time-to-frequency transformation on the DI data acquired within a slot of the S time slots, wherein the performing is applied to the DI data acquired from each transducer to obtain a plurality of transducer complex frequency spectra;

calculating a cross-channel spectral density matrix (CSDM) from the plurality of transducer complex frequency spectra, wherein the CSDM is calculated as a time average over a plurality of heart cycles;

performing an eigenvalue-eigenvector decomposition (EED) of the CSDM to obtain a spatial frequency number verses temporal frequency representation of the DI data.

37. The apparatus of claim 36, wherein the method further comprises:

computing an Equivalent Rectangular Bandwidth (ERB) of the DI data for the p-th spatial eigenvalue, wherein the ERB=C_s(p)*(FFT frequency resolution), and C_s(p) is a counter used during the summation λ

_s(p).