Detection/exclusion of acute myocardial infarction using neural network analysis of measurements of biochemical markers

• US 5,690,103 A
• Filed: 06/20/1996
• Issued: 11/25/1997
• Est. Priority Date: 06/20/1996
• Status: Expired due to Fees

• Current Assignee: Johan Ellenius, Torgny Lars Groth
• Original Assignee: Johan Ellenius, Torgny Lars Groth
• Inventors: Ellenius, Johan, Groth, Torgny Lars
• Primary Examiner(s): Kamm, William E.
• Application Number:
  US08/668,260
• Time in Patent Office: 523 Days
• Field of Search: 128/632, 128/924
• US Class Current: 600/368
• CPC Class Codes: G01N 2800/324 Coronary artery diseases, e.g. angina pectoris, myocardial infarction
• CPC Class Codes: G01N 33/6887 from muscle, cartilage or connective tissue
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: Y10S 128/924 using artificial intelligence

Abstract

The overall invention categorizes patients with suspected acute myocardial infarction (AMI) with regard to a) AMI/non-AMI; b) infarct size (e.g. Major/Minor); c) time since onset of infarction; and d) non-AMI with/without minor myocardial damage (MMD). Generally, the above categorization is based on frequent timed blood sampling and measurement of selected biochemical markers of AMI with different rates of appearance in circulating blood. The computations are performed by using specially designed artificial neural networks. According to a first main aspect of the invention, early, i.e. generally within 3 hours from admission of the patient, detection/exclusion of acute myocardial infarction is provided. Furthermore, early prediction of the infarct size and early estimation of the time from onset are also provided.

Claims

1. A method for early detection of acute myocardial infarction (AMI) in a patient, comprising the steps of:

measuring, with a predetermined timing and at at least two different time instances occurring within 3 hours from admission of said patient, concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

generating an indication of AMI based on a first predetermined set of said concentration levels and first empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients by using a neural network structure trained with said first empirical data to detect marker-patterns of AMI; and

predicting the infarct size based on a second predetermined set of said concentration levels and second empirical data comprising concentration levels obtained from previous patients with different infarct sizes and a preclassification of the infarct sizes of said previous patients, whereby an early detection of AMI and an early infarct size prediction are obtained.

2. A method according to claim 1, wherein said first predetermined set comprises concentration levels measured at at least one of said at least two time instances, and said second predetermined set comprises concentration levels measured at more than one of said at least two time instances.

3. A method according to claim 1, wherein said at least two biochemical markers are selected from the following group consisting of:

myoglobin, creatine kinase MB, troponin-T, troponin-I, glycogen phosphorylase BB, lactate dehydrogenase, and heart-type fatty acid binding protein (h-FABP).

4. A method according to claim 1, wherein said predetermined timing generally corresponds to the timing with which the concentration levels of said first and second empirical data were obtained.

5. A method according to claim 1, wherein said neural network structure comprises at least two neural networks, each one of which is trained to detect AMI within an individual defined period of time after onset of symptoms of infarction for generating an individual indication of AMI/non-diagnosis, said individual indications being used for generating said indication of AMI.

6. A method according to claim 1, wherein said neural network structure comprises a feedback connection for recognizing temporal patterns of the concentration levels of said first predetermined set, said first predetermined set comprising concentration levels measured at more than one of said at least two time instances.

7. A method according to claim 1, wherein said predicting step comprises the step of using a further neural network structure trained with said second empirical data to detect marker-patterns of different infarct sizes.

8. A method according to claim 1 further comprising the step of generating, based on a third predetermined set of said concentration levels and third empirical data comprising concentration levels obtained from previous patients of different states regarding the time after onset of infarction and a preclassification of the times after onset for said previous patients, at least one time estimate representing a time period within which onset of infarction occurred.

9. A method according to claim 8, wherein said step of generating at least one time estimate comprises the step of using a third neural network structure trained with said third empirical data to detect marker-patterns of times after onset of infarction.

10. A method according to claim 9, wherein at least two time estimates are generated, said method further comprising the step of sending said at least two time estimates to a decision table for generating a final estimate of a time period within which said onset of infarction occurred.

11. A method according to claim 1, wherein said step of generating an indication of AMI further comprises the step of preprocessing said concentration levels, and wherein said preprocessed concentration levels are processed by said neural network structure.

12. An apparatus for early detection of acute myocardial infarction (AMI) in a patient, comprising:

means for performing timed measurements, at at least two different time instances occurring within 3 hours from admission of said patient, of concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

a neural network structure for generating an indication of AMI based on a first predetermined set of said concentration levels and first empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients, said neural network structure being trained with said first empirical data to detect marker-patterns of AMI; and

means for predicting the infarct size based on a second predetermined set of said concentration levels and second empirical data comprising concentration levels obtained from previous patients with different infarct sizes and a preclassification of the infarct sizes of said previous patients.

13. An apparatus according to claim 12, wherein said neural network structure comprises at least two neural networks, each one of which is associated with a separate period of time after onset of symptoms of infarction for generating an individual indication of AMI/non-diagnosis;

and wherein said apparatus further comprises means for generating said indication of AMI based on said individual indications of AMI/non-diagnosis.

14. An apparatus according to claim 12, wherein said neural network structure is a two-layer neural network with a feedback connection for recognizing temporal patterns of the concentration levels of said first predetermined set so as to generate said indication of AMI.

15. An apparatus according to claim 12, wherein said predicting means comprises a further neural network structure trained with said second empirical data to detect marker-patterns of different infarct sizes.

16. An apparatus according to claim 12 further comprising means for generating, based on a third predetermined set of said concentration levels and third empirical data comprising concentration levels obtained from previous patients of different states regarding the time after onset of infarction and a preclassification of the times after onset for said previous patients, at least one time estimate representing a time period within which onset of infarction occurred.

17. An apparatus according to claim 16, wherein said means for generating at least one time estimate comprises a neural network structure trained with said third empirical data to detect marker-patterns of times after onset of infarction.

18. An apparatus according to claim 12 further comprising a unit for preprocessing said concentration levels, and wherein said preprocessed concentration levels are processed by said neural network structure.

19. A method for early detection of acute myocardial infarction (AMI) in a patient, comprising the steps of:

performing timed measurements, at at least two different time instances occurring within 3 hours from admission of said patient, of concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

generating an indication of AMI based on a first predetermined set of said concentration levels and first empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients by using a neural network structure trained with said first empirical data to detect marker-patterns of AMI;

predicting the infarct size based on a second predetermined set of said concentration levels and second empirical data comprising concentration levels obtained from previous patients with different infarct sizes and a preclassification of the infarct sizes of said previous patients; and

generating, based on a third predetermined set of said concentration levels and third empirical data comprising concentration levels obtained from previous patients of different states regarding the time after onset of infarction and a preclassification of the times after onset for said previous patients, at least one time estimate representing a time period within which onset of infarction occurred,

whereby an early detection of AMI, an early infarct size prediction and an early estimate of the time after onset of AMI are obtained.

20. A method for:

a) early detection/exclusion of acute myocardial infarction (AMI) in a patient;

b) predicting the infarct size;

c) estimating the time after onset of infarction;

d) assessing/excluding possible minor myocardial damage (MMD), said method comprising the steps of;

measuring, with a predetermined timing and at at least two different time instances occurring within 3 hours from admission of said patient, concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

generating an indication of AMI/non-diagnosis by making a comparison based on a first predetermined set of said concentration levels and first empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients by using a neural network structure trained with said first empirical data to detect marker-patterns of AMI;

predicting the infarct size by making a comparison based on a second predetermined set of said concentration levels andsecond empirical data comprising concentration levels obtained from previous patients with different infarct sizes and a preclassification of the infarct sizes of said previous patients, provided an indication of AMI has been generated;

generating, by making a comparison based on a third predetermined set of said concentration levels and third empirical data comprising concentration levels obtained from previous patients of different states regarding the time after onset of infarction and a preclassification of the times after onset for said previous patients, an estimate representing a time period within which onset of infarction occurred, provided an indication of AM/has been generated;

excluding AMI provided said indication of AMI has not been generated for a predetermined period of time; and

assessing/excluding MMD by making a comparison based on a fourth predetermined set of said concentration levels and fourth empirical data comprising concentration levels obtained from previous patients with MMD/without MMD and a preclassification of the presence/absence of MMD in said previous patients, provided AMI has been excluded.

21. A method for early detection of acute myocardial infarction (AMI) in a patient, comprising the steps of:

measuring, with a predetermined timing and at at least two different time instances occurring within 3 hours from admission of said patient, concentration levels of least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

generating an indication of AMI based on a first predetermined set of said concentration levels and first empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients by using a neural network structure trained with said first empirical data to detect marker-patterns of AMI; and

generating, based on a third predetermined set of said concentration levels and third empirical data comprising concentration levels obtained from previous patients of different states regarding the time after onset of infarction and a preclassification of the times after onset for said previous patients, an estimate of a time period within which onset of infarction occurred,

whereby an early detection of AMI and an early estimate of the time after onset of AMI are obtained.

22. An apparatus for excluding acute myocardial infarction (AMI) in a patient, comprising:

means for measuring, with a predetermined timing and within 3 hours from admission of said patient, concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

means for generating an indication of AMI based on a first predetermined set of said concentration levels and first empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients by using a neural network structure trained with said first empirical data to detect marker-patterns of AMI; and

means for excluding AMI as a diagnosis provided said indication of AMI has not been generated for a predetermined period of time.

23. Apparatus according to claim 22 further comprising means for assessing/excluding minor myocardial damage (MMD) based on said concentration levels and fourth empirical data comprising concentration levels obtained from previous patients with MMD/without MMD and a preclassification of the presence/absence of MMD in said previous patients, provided AMI has been excluded as a diagnosis.

24. An apparatus according to claim 23, wherein said means for assessing/excluding MMD comprises a fourth neural network structure trained with said fourth empirical data to detect marker-patterns of MMD, said neural network structure comprising a feedback connection for recognizing temporal patterns of said concentration levels.

25. A method for excluding acute myocardial infarction (AMI) as a diagnosis for a patient, comprising the steps of:

performing timed measurements, at at least two different time instances occurring within 3 hours from admission of said patient, of concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

generating an indication of non-diagnosis based on a first predetermined set of said concentration levels and first empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said patients, by using a neural network structure trained with said first empirical data to detect marker-patterns of AMI;

excluding AMI as a diagnosis provided said indication of non-diagnosis has been generated for a predetermined period of time; and

assessing/excluding minor myocardial damage (MMD) based on said concentration levels and fourth empirical data comprising concentration levels obtained from previous patients with MMD/without MMD and a preclassification of the presence/absence of MMD in said previous patients, provided AM/has been excluded as a diagnosis,

whereby an early exclusion of AMI and an assessment/exclusion of MMD are obtained.

26. A method according to claim 25, wherein said step of assessing/excluding MMD comprises the step of using a neural network structure trained with said fourth empirical data to detect marker-patterns of MMD.

27. A method according to claim 26, wherein said neural network structure comprises a feedback connection for recognizing temporal patterns of said concentration levels.

28. An apparatus for early detection of acute myocardial infarction (AMI) in a patient, comprising:

means for measuring concentration levels of at least two biochemical markers of AMI in plasma of said patient and for generating signals representing said concentration levels, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

a neural network structure responsive to said signals and trained with empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients to detect marker-patterns of AMI, said neural network structure comprising at least two neural networks, each one of which is trained to detect AMI within an individual defined time period after onset of symptoms of infarction for generating an individual indication of AMI/non-diagnosis; and

means for generating a conclusive indication of AMI based on said individual indications.

29. An apparatus according to claim 28, further comprising means for generating an estimate representing a time period within which onset of infarction occurred based on said individual defined time periods and said individual indications.

30. An apparatus for early detection of acute myocardial infarction (AMI) in a patient, comprising:

means for measuring concentration levels of at least two biochemical markers of AMI in plasma of said patient and for generating signals representing said concentration levels, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

a neural network structure responsive to said signals and trained with empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients to detect marker-patterns of AMI, said neural network structure comprising a feedback connection for recognizing temporal patterns of said concentration levels so as to generate an indication of AMI.

31. An apparatus for early detection of acute myocardial infarction (AMI) in a patient, comprising:

means for measuring concentration levels of at least two biochemical markers of AMI in plasma of said patient and for generating signals representing said concentration levels, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

a first neural network structure responsive to the initial signals and trained with first empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous patients to detect marker-patterns of AMI, said first neural network structure comprising at least two neural networks, each one of which is associated with a separate period of time after onset of symptoms of infarction for generating an individual indication of AMI/non-diagnosis;

means for generating a first indication of AMI/non-diagnosis based on said individual indications; and

a second neural network structure responsive to said signals and trained with said first empirical data to detect marker-patterns of AMI, said second neural network structure comprising a feedback connection for recognizing temporal patterns of said concentration levels so as to generate a second indication of AMI provided said first indication of AMI was not generated.

32. A method for early detection of acute myocardial infarction (AMI) in a patient, comprising the steps of:

measuring concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI;

generating an indication of AMI based on a first predetermined set of said concentration levels and empirical data comprising concentration levels obtained from previous patients with AMI/non-AMI and a preclassification of the presence/absence of AMI in said previous by using a neural network trained with said first empirical data to detect marker-patterns of AMI, the training of said neural network being performed by maximizing the net gain of weighted true and false classifications.

33. A method for predicting infarct size for a patient with acute myocardial infarction (AMI), comprising the steps of:

measuring concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI; and

predicting the infarct size based on said concentration levels and empirical data comprising concentration levels obtained from previous patients with different infarct sizes and a preclassification of the infarct sizes of said previous patients by using a neural network trained with said empirical data to detect marker-patterns of different infarct sizes.

34. A method for assessing minor myocardial damage (MMD) in a patient, comprising the steps of:

measuring concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI; and

assessing/excluding MMD based on said concentration levels and empirical data comprising concentration levels obtained from previous patients with MMD/without MMD and a preclassification of the presence/absence of MMD in said previous patients by using a neural network structure trained with said empirical data to detect marker-patterns of MMD.

35. A method for estimating the time after onset of infarction in a patient with acute myocardial infarction, comprising the steps of:

measuring concentration levels of at least two biochemical markers of AMI in plasma of said patient, said at least two biochemical markers having different rates of appearance in said plasma at AMI; and

generating an estimate representing a time period within which onset of infarction occurred by making a comparison based on said concentration levels and empirical data comprising concentration levels obtained from previous patients of different states regarding the time after onset of infarction and a preclassification of the times after onset of infarction for said previous patients.