Method and apparatus for predicting patient outcomes from a physiological segmentable patient signal

• US 8,346,349 B2
• Filed: 01/15/2009
• Issued: 01/01/2013
• Est. Priority Date: 01/16/2008
• Status: Active Grant

• Current Assignee: Massachusetts Institute of Technology
• Original Assignee: Massachusetts Institute of Technology
• Inventors: Guttag, John V., Syed, Zeeshan H., Sung, Philip Pohong, Stultz, Collin M.
• Primary Examiner(s): Layno, Carl H
• Assistant Examiner(s): Stice, Paula J
• Application Number:
  US12/321,239
• Publication Number:
  US 20090192394A1
• Time in Patent Office: 1,447 Days
• Field of Search: 600509-525
• US Class Current: 600/509
• CPC Class Codes: A61B 5/02405 Determining heart rate variability
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/349 Detecting specific parameters of the electrocardiograph cycle
• CPC Class Codes: A61B 5/363 Detecting tachycardia or bradycardia
• CPC Class Codes: A61B 5/7275 Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

A method and apparatus for predicting patient outcome from a physiological segmentable signal of a patient. In one embodiment, the method comprises the steps of obtaining the physiological segmentable signal of the patient; segmenting the physiological segmentable signal into a plurality of separate segmentable components; calculating a time series of the morphological distance between adjacent separate segmentable components of the plurality of separate segmentable components; and predicting patient outcome in response to the time series of the morphological distance. In another aspect, the invention relates to a method for extracting information from physiological signals for one or more subjects including the steps of partitioning the physiological signal into a plurality of components, grouping the components into a plurality of information classes, assigning a unique symbol to each information class, mapping each component to the assigned symbol, and examining one or more such sequences for clinical significance.

Claims

1. A method of predicting patient outcome from a physiological segmentable signal of a patient comprising the steps of:

obtaining the physiological segmentable signal from a medical device in communication with the patient;

segmenting, using a processor, the physiological segmentable signal into a plurality of separate components;

calculating, using a processor, a time series of a morphological distance between adjacent separate components of the plurality of separate components,

the morphologic distance between adjacent separate components being determined by an energy difference in the adjacent separate components, and

the energy difference between adjacent separate components being determined by physiological dynamic time warping; and

predicting, using a processor, patient outcome at least in response to the time series of the morphological distance.

2. The method of claim 1 wherein the physiological segmentable signal is an ECG and where each segmentable component is a heartbeat.

3. The method of claim 2 wherein segments of the ECG that are unlikely to represent cardiac activity are rejected.

4. The method of claim 2 wherein the step of predicting patient outcome is performed with an overall risk measure calculated in response to an ECG risk measure and a plurality of other metrics.

5. The method of claim 1 wherein a higher variability of morphologic distance indicates a higher probability of a worse patient outcome.

6. The method of claim 1 wherein the predicted patient outcome is selected from a group consisting of death, arrhythmia, myocardial ischemia and myocardial infarction.

7. The method of claim 1 wherein the method further comprises the step of predicting patient outcome in response to combining, using a processor, morphologic distance variability with a clinical predictor.

8. The method claim 7 wherein the other clinical predictor is a TIMI risk assessment.

9. The method of claim 1 wherein the step of predicting patient outcome is performed with a risk measure calculated in response to morphological distance variability and other measures.

10. The method of claim 1 further comprising the steps of noise and artifact removal.

11. The method of claim 1 wherein the physiological signal is a respiration signal.

12. The method of claim 11 wherein each component is a breath.

13. The method of claim 1 wherein an information class is derived from at least one of:

standard features of clinical significance and non-standard features of clinical significance.

14. The method of claim 1 wherein the components are grouped into information classes using clustering.

15. The method of claim 1 wherein the plurality of components is grouped into a predetermined number of information classes.

16. The method of claim 1 wherein the plurality of components is grouped into a number of information classes which is not predetermined.

17. The method of claim 1 further comprising the step of associating a representative component to an information class.

18. The method of claim 17 wherein the information class is formed using a clustering technique and the representative component of the information class is a centroid of the cluster.

19. The method of claim 1 further including the steps of

repeating the steps for a plurality of patients having a common components of interest;

assigning a symbol to each of the components of interest to for a symbolic series and

examining the symbolic series for commonalities of the common components of interest.

20. The method of claim 19 wherein a commonality is a pattern that is approximately conserved preceding the event of interest.

21. The method of claim 19 wherein commonality is determined using active regions and a two-layer Gibbs sampling algorithm.

22. The method of claim 19 wherein both positive and negative examples are used to determine commonalities.

23. The method of claim 1 wherein an information class is derived from information in a time and/or frequency domain of the physiological segmentable signal.

24. The method of claim 23 wherein the information class is assigned a symbol and wherein a sequence of symbols has significance due to at least one of:

the number of distinct symbols represented and the distribution of symbols represented.

25. The method of claim 23 wherein the information class is assigned a symbol and wherein a sequence of symbols has significance due to its entropy.

26. The method of claim 23 wherein the information class is assigned a symbol and wherein a sequence of symbols has significance due to the pattern of symbols represented.

27. The method of claim 24 in which an examination of the symbolic series is performed on signals from a group consisting of signals from different time periods for the same patient and signals from different patients.

28. The method of claim 24 in which an examination of the symbolic series is performed using methods selected from the group of locality sensitive hashing, clustering using a 2-approximate solution of the k-center problem and sequential statistics.

29. The method of claim 24 wherein the examination of the symbolic sequences is based upon comparing patients to other patients with related symbolic sequences.

30. The method of claim 24 further comprising associating a representative component to each information class and wherein the matching is performed based upon characteristics of a group consisting of the differences among the representative components of the information classes, the distribution of symbols, and hidden Markov models.

31. The method of claim 27 wherein patients are partitioned into a dominant group and a plurality of non-dominant groups.

32. The method of claim 27 wherein a new patient is assigned a risk profile derived from matching the new patient'"'"'s physiological signal to groups derived from historical data.

33. The method of claim 31 wherein patients in non-dominant groups are assessed as being at elevated risk.

34. A method of predicting patient outcome from a physiological segmentable signal of a patient comprising the steps of:

obtaining the physiological segmentable signal from a medical device in communication with the patient;

segmenting, using a processor, the physiological segmentable signal into a plurality of separate components;

calculating, using a processor, a time series of a morphological distance between adjacent separate components of the plurality of separate components,

the morphologic distance between adjacent separate components being determined by an energy difference in the adjacent separate components, and

the energy difference between adjacent separate components being determined by physiological dynamic time warping; and

summarizing, using a processor, the information in the time series.

35. The method of claim 34 further comprising the step of predicting patient outcome in response to the summarized time series information.

36. The method of claim 34 further comprising treating the patient by a clinician and measuring efficacy of treatment in response to the summarized time series information by the processor.

37. The method of claim 34 wherein the step of summarizing of time series information comprises the steps of determining the low frequency power and high frequency power in the time series and calculating a ratio of low frequency power to high frequency power for the time series.

38. The method of claim 34 wherein the summarizing of the time series information comprises the steps of determining the low frequency power and high frequency power in the time series and calculating energy in a diagnostic frequency range.

39. The system of claim 34 wherein an information class is derived from information in one of a time and frequency domain of the physiological segmentable signal.

40. The system of claim 39 wherein each information class is assigned a symbol and wherein the sequence of symbols has significance due to at least one of:

the number of distinct symbols represented and the distribution of symbols represented.

41. The system of claim 40 wherein the symbolic sequence has significance due to a group consisting of its entropy and the pattern of symbols represented.

42. The system of claim 34 wherein an information class is one of standard clinical significance and non-standard clinical significance.

43. The system of claim 34 wherein the components are grouped into information classes using clustering.

44. The system of claim 34 wherein the plurality of components is grouped into a predetermined number of information classes.

45. The system of claim 34 wherein the plurality of components is grouped into a number of information classes which is not predetermined.

46. The system of claim 34 further comprising associating a representative component to each information class.

47. A system for predicting patient outcome from a physiological segmentable signal of a patient comprising:

a processor, the processor performing the operations of;

obtaining the physiological segmentable signal of the patient;

segmenting the physiological segmentable signal into a plurality of separate components; and

calculating a time series of a morphological distance between adjacent separate components of the plurality of separate components,

the morphologic distance between adjacent separate components being determined by an energy difference in the adjacent separate components,

the energy difference between adjacent separate components being determined by physiological dynamic time warping; and

predicting patient outcome in response to the time series.

48. The system of claim 47 wherein the processor further predicts patient outcome in response to the time series of the morphological distance.

49. The system of claim 47 wherein the processor, after the patient has been treated by a clinician, further performs the step of determining the efficacy of treatment in response to the time series of morphological distance.

50. The system of claim 47 wherein the physiological segmentable signal is an ECG and where each component is a heartbeat.

51. The system of claim 47 wherein the higher the variability of morphologic distance indicates a higher probability of a worse the patient outcome.

52. A system for predicting patient outcome from a physiological segmentable signal of a patient comprising

a processor, the processor performing the operations of:

obtaining the physiological segmentable signal of the patient;

segmenting the physiological segmentable signal into a plurality of separate components;

calculating a time series of a morphological distance between adjacent separate components of the plurality of separate components,

the morphologic distance between adjacent separate components being determined by the energy difference in the adjacent separate components, and

the energy difference between adjacent separate components being determined by physiological dynamic time warping; and

summarizing the information in the time series.

53. The system of claim 52 wherein the processor further performs the step of predicting patient outcome in response to the summarized time series information.

54. The system of claim 52 wherein the processor, after the patient has been treated by a clinician, further performs the step of determining efficacy of treatment in response to the summarized time series information.

55. The system of claim 52 wherein the physiological signal is an ECG signal.

56. The system of claim 55 wherein the each component is a heartbeat.

57. The system of claim 52 wherein the physiological signal is a respiration signal.

58. The system of claim 57 wherein the each component is a breath.