Method of biomedical signal analysis including improved automatic segmentation

• US 8,332,017 B2
• Filed: 11/30/2007
• Issued: 12/11/2012
• Est. Priority Date: 12/01/2006
• Status: Expired due to Fees

• Current Assignee: Obs Medical Limited
• Original Assignee: Obs Medical Limited
• Inventors: Tarassenko, Lionel, Patterson, Alan, Strachan, Iain Guy
• Primary Examiner(s): Gedeon, Brian T
• Application Number:
  US12/517,127
• Publication Number:
  US 20100056939A1
• Time in Patent Office: 1,838 Days
• Field of Search: 600/509, 600/523, 704/242
• US Class Current: 600/509
• CPC Class Codes: A61B 5/349 Detecting specific parameters of the electrocardiograph cycle
• CPC Class Codes: A61B 5/36 Detecting PQ interval, PR interval or QT interval
• CPC Class Codes: A61B 5/7203 for noise prevention, reduction or removal
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

A method of analysing biomedical signals, for example electrocardiograms, by using a Hidden Markov Model for subsections of the signal. In the case of an electrocardiogram two Hidden Markov Models are used to detect respectively the start and end of the QT interval. The relationship between the QT interval and heart rate can be computed and a contemporaneous value for the slope of this relationship can be obtained by calculating the QT/RR relationship for all of the beats in a sliding time window based on the current beat. Portions of electrocardiograms taken on different days can efficiently and accurately be compared by selecting time windows of the ECGs at the same time of day, and looking for similar beats in those time windows.

Claims

1. A computer-implemented method for analysis of an electrocardiogram, the method comprising segmenting the electrocardiogram into different segments corresponding to different cardiac features using a Hidden Markov Model, the model comprising a plurality of states corresponding to the different cardiac features;

wherein;

the segmenting is performed using two or more Hidden Markov Models corresponding only to a subset of the cardiac features the subset comprising the QRS co ex and the T wave within the QT interval of an electrocardiogram, the QT interval being the period from Q_onset to T_offset in the electrocardiogram,

the segmenting to identify the T wave bein erformed based on the estimated position of T_offset as the end point of the QT interval, and

wherein the position of the endpoint of the QT interval is obtained from a relationship between heart rate and QT interval.

2. The method according to claim 1, wherein a confidence measure is calculated based on a difference between the position of the start or end point as estimated by reference to the R peak or J point of the electrocardiogram and as obtained from the segmenting.

3. The method according to claim 1, wherein a confidence measure is calculated based on a probability of the segmenting obtained from the HMM.

4. The method according to claim 2, wherein the confidence measure is used to reject automatically artifacts in the electrocardiogram.

5. The method according to claim 1, wherein the relationship between heart rate and QT interval is obtained by analysis of the electrocardiogram over a period preceding the beat being segmented.

6. The method according to claim 1, further comprising correcting variations in the QT interval in the electrocardiogram caused by variations in heart rate, the method further comprising finding values of QT interval and corresponding beat to beat time for each of a plurality of beats in a time period spanning a current beat, the beat to beat time being the time period between the R peak of each beat and the R peak of a neighbouring beat, performing regression analysis on the QT intervals and corresponding beat to beat times over the time period to calculate a relationship between the QT interval and beat to beat time, and using that relationship to correct QT for current beat, wherein the regression analysis is periodically repeated based on a new time period spanning the current beat at that time.

7. The method according to claim 6, wherein the regression analysis is repeated for each beat based on the time period spanning each respective beat.

8. The method according to claim 6, wherein the time period spanning the current beat is the time period from 2 or 3 hours before to 2 or 3 hours after the current beat.

9. The method according to claim 1, further comprising finding, for each of a plurality of beats in the electrocardiogram a current value of the slope of the relationship between the QT interval and R-R interval, the R-R interval being the time period between the R peak of the corresponding beat and the R peak of a neighbouring beat, each current value of the slope being obtained by performing regression analysis on values of QT intervals and corresponding R-R intervals for beats in a respective time period spanning each of the plurality of beats, the method further comprising displaying the calculated current slope values for each of the plurality of beats, thereby to display variations in the slope through the electrocardiogram.

10. The method according to claim 9, wherein the respective time periods are each the time period of from 4 to 6 hours spanning each respective one of the plurality of beats.

11. The method according to claim 1, further comprising analysing two electrocardiograms of a patient taken on different days, comprising selecting a portion of each of the electrocardiograms obtained during the same time period on each of the different days, defining a similarity metric on the basis of one or more features of the electrocardiograms to measure similarity of subsections of the two portions, using the similarity metric to select a subsection of each of the two portions which are similar, and performing comparative analysis on the selected subsections to determine changes in a feature of the electrocardiograms different from the one or more features of the signal.

12. The method according to claim 11, wherein the step of selecting a portion of each of the electrocardiograms obtained during the same time period on each of the different days comprises selecting a time of day and setting as the time period a time window relative to that time point.

13. The method according to claim 11, further comprising selecting a plurality of times of day and setting a time window relative to each of them.

14. The method according to claim 11, wherein the similarity metric measures shape of the waveform of the electrocardiograms.

15. The method according to claim 11, wherein the comparative analysis is QT interval measurement.

16. The method according to claim 11, wherein the similarity metric measures the heart rate.

17. The method according to claim 11, wherein the subsections selected for comparative analysis are several of the most similar individual beats in the portions of the electrocardiograms.

18. The method according to claim 17, wherein three beats are selected.

19. The method according to claim 11, wherein the comparative analysis is on the basis of a value of a feature of the electrocardiograms averaged over each of the two subsections.