Heart monitoring apparatus and method

• US 5,749,367 A
• Filed: 09/05/1996
• Issued: 05/12/1998
• Est. Priority Date: 09/05/1995
• Status: Expired due to Term

• Current Assignee: Cardionetics Limited
• Original Assignee: Cardionetics Limited
• Inventors: Needham, Philip, Gamlyn, Lee, O'Sullivan, Siobhan, Harris, Tom
• Primary Examiner(s): GETZOW, SCOTT M
• Application Number:
  US08/709,539
• Time in Patent Office: 614 Days
• Field of Search: 128/696, 128/702, 128/699, 369/413.06
• US Class Current: 600/509
• CPC Class Codes: A61B 5/0006 ECG or EEG signals
• CPC Class Codes: A61B 5/349 Detecting specific parameters of the electrocardiograph cycle
• CPC Class Codes: A61B 5/7203 for noise prevention, reduction or removal
• CPC Class Codes: A61B 5/7225 Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation A61B5/30, A61B5/316 take precedence; transformers for specific applications H01F38/00; amplifiers H03F
• CPC Class Codes: A61B 5/7257 using Fourier transforms
• CPC Class Codes: A61B 5/726 using Wavelet transforms
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

A heart monitoring apparatus and method is disclosed wherein an electrocardiograph signal is obtained from a patient and processed to enhance the salient features and to suppress noise. A plurality n of values representative of the features of the electrocardiograph signal are generated and used in a Kohonen neural network to generate an n dimensional vector. This vector is compared with a stored plurality m of n dimensional reference vectors defining an n dimensional Kohonen feature map to determine the proximity of the vector to the reference vectors. If it is determined by the Kohonen neural network that the vector is within or beyond a threshold range of the reference vectors a signal is output which can be used to initiate an event such as the generation of an alarm or the storage of ECG data.

Claims

1. Heart monitoring apparatus comprising:

detection means for obtaining an electrocardiograph signal from a patient during a monitoring phase,

preprocessing means for processing said electrocardiograph signal to enhance the salient features of the electrocardiograph signal and suppress the noise, and to generate a plurality n of values representative of the features of said electrocardiograph signal,

storage means for storing a plurality m of n dimensional reference vectors;

Kohonen neural network means for receiving said plurality of values during the monitoring phase, for forming an n dimensional vector from said plurality of values and for comparing said n dimensional vector with said stored plurality m of n dimensional reference vectors defining an n dimensional Kohonen feature map to determine the proximity of said n dimensional vector to said reference vectors, and

output means for outputting a signal if said Kohonen neural network means determines that said n dimensional vector is within or beyond a threshold range of said reference vectors.

2. Heart monitoring apparatus as claimed in claim 1 wherein said Kohonen neural network means includes means to receive, during a learning phase, a plurality of reference values representative of the features of a reference electrocardiograph signal, means to generate an n dimensional Kohonen feature map having said plurality m of reference vectors, and means to store said reference vectors in said storage means.

3. Heart monitoring apparatus as claimed in claim 2 wherein said detection means includes means to obtain the reference electrocardiograph signal from the patient during the learning phase;

said Kohonen neural network means includes means to, during the monitoring phase, determine whether the n dimensional vector lies outside a predetermined range of said reference vectors; and

said output means includes means to output said signal if said Kohonen neural network means determines that said n dimensional vector lies outside said predetermined range.

4. A heart monitoring apparatus as claimed in claim 2 including reference input means to input said reference electrocardiograph signal to said preprocessing means.

5. Heart monitoring apparatus as claimed in claim 4 wherein input means includes means to input said reference electrocardiograph signal representing normal range of operation of the heart;

said Kohonen neural network means includes means to determine whether the n dimensional vector lies outside a predetermined range of said reference vectors; and

said output means includes means to output said signal if said Kohonen neural network means determines that said n dimensional vector lies outside said predetermined range.

6. Heart monitoring apparatus as claimed in claim 4 wherein said input means includes means to input at least one abnormal said reference electrocardiograph signal representing at least one known heart condition;

said Kohonen neural network means includes means to determine whether the n dimensional vector lies within a predetermined range of the abnormal reference vectors; and

said output means includes means to output said signal if said Kohonen neural network means determines that said n dimensional vector lies inside said predetermined range.

7. Heart monitoring apparatus as claimed in claim 2 wherein said Kohonen neural network means includes means to, in the learning phase, receive a second plurality of reference values representative of features of an electrocardiograph signal indicative of a specific heart condition, to generate a plurality of abnormal reference vectors in said n dimensional Kohonen feature map;

and means to in the monitoring phase, compare said n dimensional vector with said plurality of abnormal reference vectors to determine whether said n dimensional vector lies within a predetermined range of said abnormal reference vectors;

said apparatus including heart condition warning means responsive to said output signal for generating a warning that said patient has said specific heart condition if said Kohonen neural network means determines that said n dimensional vector lies within said predetermined range of said abnormal reference vectors.

8. Heart monitoring apparatus as claimed in claim 1 including vector inputting means for inputting said plurality m of reference vectors defining said n dimensional Kohonen feature map to said storage means.

9. Heart monitoring apparatus as claimed claim 1 wherein said preprocessing means includes means to carry out feature extraction to extract important features from said electrocardiograph signal, and to use said features to form said plurality n of values.

10. Heart monitoring apparatus as claimed in claim 9 wherein said preprocessing means includes means to identify the peaks in said electrocardiograph signal and to measure the peak values to form said plurality n of values.

11. Heart monitoring apparatus as claimed in claim 10 wherein said preprocessing means includes means to identify points midway between peaks and to measure the values of the electrocardiograph signal at said points to include the values in said plurality n of values.

12. Heart monitoring apparatus as claimed in claim 11 wherein said preprocessing means includes means to identify the P, Q, R, S and T peaks and mid points between the P and Q peaks and the S and T peaks in said electrocardiograph signal, and to use the values of the signal at the identified peaks and points to form said plurality n of values.

13. Heart monitoring apparatus as claimed in claim 1 wherein preprocessing means includes means to perform a transform of said electrocardiograph signal.

14. Heart monitoring apparatus as claimed in claim 13 wherein said preprocessing means includes means to perform a Fourier transform and/or a Wavelett transform of said electrocardiograph signal.

15. Heart monitoring apparatus as claimed in claim 1 wherein said Kohonen neural network means includes means to carry out said comparison and determination by determining the differences between said reference vectors and said n dimensional vector for each dimension, summing the individual dimensional differences and comparing the sum with a threshold difference value.

16. Heart monitoring apparatus as claimed in claim 1 wherein said Kohonen neural network means includes means to carry out said comparison and determination by mapping said n dimensional vector into said Kohonen feature map, calculating the differences therebetween by Pythagoras'"'"' law, and comparing said differences with at least one difference threshold.

17. Heart monitoring apparatus as claimed in claim 1 including filter means to remove distinctive irregular heartbeats from the electrocardiograph signal.

18. Heart monitoring apparatus as claimed in claim 17 including indicating means to indicate the occurrence of the distinctive irregular beats.

19. Heart monitoring apparatus as claimed in claim 18 wherein said indicating means comprises a low pass filter.

20. Heart monitoring apparatus as claimed in claim 17 or claim 17 wherein said filter means comprises a high pass filter.

21. Heart monitoring apparatus as claimed in claim 17 wherein said storage means stores a first set of n dimensional reference vectors for the identification of the distinctive irregular heartbeat, and a second set of n dimensional reference vectors for monitoring regular heartbeats;

said Kohonen neural network means includes means to, during the monitoring phase, initially compare the n dimensional vector with the stored first set of reference vectors to identify the distinctive irregular heartbeats, and to subsequently compare an n dimensional vector formed from regular heartbeats which exclude the distinctive irregular heartbeats with the second set of n dimensional reference vectors; and

said output means includes means to output said signal if said Kohonen network means determines that said n dimensional vector formed from said regular heartbeats is within or beyond a threshold range of said second set of reference vectors.

22. Heart monitoring apparatus as claimed in claim 21 wherein said output means includes means to output a signal indicating the occurrence of a distinctive irregular heartbeat when said Kohonen neural network means determines that the n dimensional vector formed from the distinctive irregular heartbeat is within or beyond a threshold range of said first set of reference vectors.

23. Heart monitoring apparatus as claimed in claim 22 including averaging means responsive to the signal indicating the occurrence of a distinctive irregular heartbeat for averaging a plurality of said electrocardiograph signals or said plurality n of values for a plurality of regular heartbeats.

24. Heart monitoring apparatus as claimed in claim 23 including normalising means for normalising the averaged plurality of values by reducing the values by the mean values and dividing by the standard deviation.

25. Heart monitoring apparatus as claimed in claim 1 including averaging means for averaging a plurality of said electrocardiograph signals or said plurality n of values after removal of said distinctive irregular beats by said filter means.

26. Heart monitoring apparatus as claimed in claim 1 including alarm generating means for generating an alarm in response to the output signal of said output means.

27. Heart monitoring apparatus as claimed in claim 1 including data storage means for storing the electrocardiograph signal and/or the plurality of values over a period of time in response to the output signal of said output means.

28. Heart monitoring apparatus as claimed in claim 1 including postprocessing means to further process the output signal of said output means, wherein the output signal comprises an error vector, together with further data related to heart operation to provide an indication of heart stress.

29. Heart monitoring apparatus as claimed in claim 28 including means for providing said further data comprising at least one of heart rate, heart rate variation, rate of occurrence of distinctive irregular heartbeats such as ectopic beats, and the difference between the electrocardiograph signal value midway between the S and T peaks and the electrocardiograph signal value midway between the P and Q peaks.

30. Heart monitoring apparatus as claimed in claim 28 wherein said post processing means includes neural network means.

31. Heart monitoring apparatus as claimed in claim 30 wherein said neural network means comprise multi-layered perception means.

32. Heart monitoring apparatus as claimed in claim 1 including digitizing means for digitizing said electrocardiograph signal for input to said preprocessing means.

33. Heart monitoring apparatus as claimed in claim 1 wherein said detection means and said output means are remote from one another, the apparatus including communication means for airborne communications therebetween.

34. Heart monitoring apparatus as claimed in claim 33 wherein said detection means is provided in a portable housing arranged for attachment to the chest of said patient, said portable housing including transmitter means for transmitting signals to said output means;

said output means being provided in a base station, said base station including receiver means to receive the transmitted signals.

35. Heart monitoring apparatus as claimed in claim 34 wherein said transmission means and said reception means are arranged to respectively transmit and receive said electrocardiograph signals, said base station including said preprocessing means, and said Kohonen neural network means.

36. Heart monitoring apparatus as claimed in claim 34 wherein said portable housing contains said preprocessing means, and said Kohonen neural network means;

and said transmission and said reception means are arranged to respectively transmit and receive signals indicative of the proximity of said n dimensional vector to said reference vectors.

37. Heart monitoring apparatus as claimed in claim 33 wherein said detection means comprises electrodes for attachment to the chest of a patient, said apparatus including a portable housing adapted to be carried by the patient, said portable housing including transmitter means for receiving signals from said electrodes and for transmitting signals to said output means;

said electrodes being connected to said housing by wires;

said output means being provided in a base station, said base station including receiver means to receive the transmitted signals.

38. Heart monitoring apparatus as claimed in claim 1 including storage means for storage of the electrocardiograph signal or said plurality n of values for a predetermined period of time in response to the signal output by said output means.

39. A heart monitoring method comprising the steps of:

obtaining an electrocardiograph signal from a patient during a monitoring phase,

preprocessing the electrocardiograph signal to enhance the salient features of the electrocardiograph signal and suppress the noise, and to generate a plurality n of values representative of the features of the electrocardiograph signal,

forming an n dimensional vector from said plurality n of values,

comparing the n dimensional vector with a stored plurality m of n dimensional reference vectors defining an n dimensional Kohonen feature map to determine the proximity of the n dimensional vector to said reference vectors, and

outputting a signal if it is determined that the n dimensional vector is within or beyond a threshold range of said reference vectors.

40. A heart monitoring method as claimed in claim 39 including the steps during a learning phase of:

receiving a plurality of reference values representative of features of a reference electrocardiograph signal,

generating an n dimensional Kohonen feature map having said plurality m of reference vectors, and

storing said reference vectors.

41. A heart monitoring method as claimed in claim 40 including the steps of:

obtaining said electrocardiograph signal from said patient during the learning phase; and

during the monitoring phase determining whether the n dimensional vector lies outside a predetermined range of said reference vectors; and

outputting said signal if it is determined that said n dimensional vector lies outside said predetermined range.

42. A heart monitoring method as claimed in claim 41 including the steps of:

during the learning phase, inputting at least one abnormal said reference electrocardiograph signal representing at least one known heart condition; and

during the monitoring phase, determining whether the n dimensional vector lies within a predetermined range of the abnormal reference vectors, and outputting said signal if it is determined that said n dimensional vector lies inside said predetermined range.

43. A heart monitoring method as claimed in claim 40 including the steps of:

during the learning phase, obtaining said reference electrocardiograph signal representative of a normal range of operation of the heart, and processing the reference electrocardiograph signal to generate said plurality of reference values, and

during the monitoring phase, determining whether the n dimensional vector lies outside a predetermined range of said reference vectors, and outputting said signal if it is determined that said n dimensional vector lies outside said predetermined range.

44. A heart monitoring method as claimed in claim 40 including the steps of:

in the learning phase, receiving a second plurality of reference values representative of features of an electrocardiograph signal indicative of a specific heart condition, and generating a plurality of abnormal reference vectors in said n dimensional Kohonen feature map; and

in the monitoring phase, comparing said n dimensional vector with said plurality of abnormal reference vectors to determine whether said n dimensional vector lies within a predetermined range of said abnormal reference vectors, and generating a warning that said patient has said specific heart condition if it is determined that said n dimensional vector lies within said predetermined range of said abnormal reference vectors.

45. A heart monitoring method as claimed in claim 39 including the step of inputting and storing said plurality m of reference vectors defining said n dimensional Kohonen feature map.

46. A heart monitoring method as claimed in claims 39 wherein the preprocessing step comprises the steps of extracting important features from said electrocardiograph signal, and generating said plurality n of values from the extracted features.

47. A heart monitoring method as claimed in claim 46 wherein the processing step includes the steps of identifying the peaks in said electrocardiograph signal, and measuring the peak values to form said plurality n of values.

48. A heart monitoring method as claimed in claim 47 wherein said preprocessing step includes the steps of identifying points midway between peaks, and measuring the values of said electrocardiograph signal at said points to include the values in said plurality n of values.

49. A heart monitoring method as claimed in claim 48 wherein said preprocessing step includes the steps of identifying the P, Q, R, S and T peaks and mid points between the P and Q and the S and T peaks in said electrocardiograph signal, and using the values for the electrocardiograph signal at the identified peaks and points to form said plurality n of values.

50. A heart monitoring method as claimed in claim 39 wherein said preprocessing step includes the step of performing a transform of said electrocardiograph signal.

51. A heart monitoring method as claimed in claim 50 wherein said preprocessing step includes the step of performing a Fourier transform and/or a Wavelett transform of said electrocardiograph signal.

52. A heart monitoring method as claimed in claim 39 wherein the comparing step includes the step of determining the differences between said reference vectors and said n dimensional vector for each dimension, summing the individual dimensional differences, and comparing the sum with a threshold difference value.

53. A heart monitoring method as claimed in claim 39 wherein the comparing step includes the steps of mapping said n dimensional vector into said Kohonen feature map, calculating the differences therebetween by Pythagoras'"'"' law, and comparing said differences with at least one difference threshold.

54. A heart monitoring method as claimed in claim 39 including the step of removing distinctive irregular heartbeats from said electrocardiograph signal.

55. A heart monitoring method as claimed in claim 54 including the step of indicating the occurrence of the distinctive irregular beats.

56. A heart monitoring method as claimed in claim 54 including the step of averaging a plurality of said electrocardiograph signals or said plurality of values after removal of said distinctive irregular heartbeats.

57. A heart monitoring method as claimed in claim 54 including the steps of during the learning phase, storing a first set of n dimensional reference vectors for the identification of the distinctive irregular heartbeat, and a second set of n dimensional reference vectors for monitoring regular heartbeats;

the comparing step comprising initially comparing the n dimensional vector with the stored first set of reference vectors to identify the distinctive irregular heartbeats, and subsequently comparing an n dimensional vector formed from regular heartbeats excluding the distinctive irregular heartbeats with said second set of n dimensional reference vectors; and

the outputting step comprising outputting said signal if it is determined that said n dimensional vector formed from said regular heartbeats is within or beyond a threshold range of said second set of reference vectors.

58. A heart monitoring method as claimed in claim 57 wherein the outputting step includes the step of outputting a signal indicating the occurrence of a distinctive irregular heartbeat when it is determined that the n dimensional vector formed from the distinctive irregular heartbeat is within or beyond a threshold range of said first set of reference vectors.

59. A heart monitoring method as claimed in claim 58 including the step of averaging a plurality of said electrocardiograph signals or said plurality n of values for a plurality of regular heartbeats.

60. A heart monitoring method as claimed in claim 59 including the step of normalising the averaged plurality of values by reducing the values by the mean values and dividing by the standard deviation.

61. A heart monitoring method as claimed in claim 39 including the step of generating an alarm in response to the output signal.

62. A heart monitoring method as claimed in claim 39 including the step of storing the electrocardiograph signal and/or the plurality of values over a period of time in response to the output signal.

63. A heart monitoring method as claimed in claim 39 including the step of post processing the output signal, wherein the output signal comprises an error vector, together with further data related to heart operation to provide an indication of heart stress.

64. A heart monitoring method as claimed in claim 63 including the step of providing the further data comprising at least one of heart rate, heart rate variation, rate of occurrence of distinctive irregular heartbeats such as ectopic beats, and the difference between the electrocardiograph signal value midway between the S and T peaks and the electrocardiograph signal value midway between the P and Q peaks.

65. A heart monitoring method as claimed in claim 63 wherein the post processing is carried out by a neural network such as the multi-layered perception.

66. A heart monitoring method as claimed in claim 39 indicating the step of digesting the electrocardiograph signal prior to preprocessing.

67. A heart monitoring method as claimed in claim 39 wherein the obtaining step is carried out at one location and the output step is carried out at a second location remote from said first location, the method including providing airborne communications between the two locations.

68. A heart monitoring method as claimed in claim 67 wherein the electrocardiograph signals are obtained by a portable unit attached to the patient, and the output signal is provided at a base station remote from said portable unit.

69. A heart monitoring method as claimed in claim 68 wherein the electrocardiograph signals are transmitted to said base station where said preprocessing step, forming step, comparing step and output step take place.

70. A heart monitoring method as claimed in claim 68 wherein the obtaining and preprocessing steps take place in said portable unit, and said plurality n of values are transmitted to said base station.

71. A heart monitoring method as claimed in claim 39 including the step of storing the electrocardiograph signal and said plurality of values for a predetermined period of time in response to said output signal.

72. Heart monitoring apparatus comprising:

input means for receiving an electrocardiograph signal from a patient;

preprocessing means for processing said electrocardiograph signal to suppress the noise and extract important features of the electrocardiograph signal to obtain an n dimensional vector comprising a plurality n of values representative of the important features of said electrocardiograph signal;

storage means for storing a plurality m of n dimensional reference vectors;

neural network means for receiving said n dimensional vector, for comparing said n dimensional vector with said stored plurality m of n dimensional reference vectors defining an n dimensional volume to determine the proximity of said n dimensional vector to said n dimensional volume, and for outputting an indication of whether said n dimensional vector lies within or beyond a threshold range of said n dimensional volume.

73. Heart monitoring apparatus as claimed in claim 72 wherein said neural network means comprises Kohonen neural network means and said reference vectors form an n dimensional Kohonen feature map.

74. Heart monitoring apparatus comprising:

measuring means for measuring the activity of a patient'"'"'s heart and outputting an electrocardiograph signal during a learning phase and a monitoring phase;

preprocessing means for processing said electrocardiograph signal to extract important features therefrom to generate an n dimensional vector comprising a plurality n of values representative of the features of the electrocardiograph signal; and

Kohonen neural network means for, during the learning phase, using said n dimensional vector to generate an n dimensional Kohonen feature map with a plurality m of reference vectors defining a normal range of electrocardiograph features for said patient, and for during the monitoring phase, comparing said n dimensional vector with said plurality of m dimensional reference vectors defining the n dimensional Kohonen feature map to determine the proximity of said n dimensional vector to said reference vectors, and to output a signal if it is determined that said n dimensional vector is within or beyond a threshold range of said reference vectors.