Decision Support for the Initial Triage of Patients with Acute Myocardial Infarction
Författare
Summary, in English
Objectives:
To develop an automated tool for the analysis of electrocardiograms (ECG) with respect to changes that make the patient a candidate for reperfusion therapy. An additional aim was to assess the influence of the tool on the ECG classifications of three interns.
Methods and Results:
An artificial neural network was trained to interpret ECGs regarding changes making the patient a candidate for reperfusion therapy. The ECG measurements used as input to the network were obtained from the measurement program of the ECG recorders. The network was trained using a database of 3000 ECGs recorded at an emergency department. In the second step three interns classified 1000 test ECGs twice at different occasions, first without and thereafter with the advice of the neural network. The gold standard of the training and test ECGs was the classification of two experienced cardiologists. The three interns showed on average a sensitivity of 68% at a specificity of 92% without the advice of the neural network and a sensitivity of 93% at a specificity of 87% with the advice. The neural network itself showed a sensitivity of 95% at a specificity of 88%. The increase in sensitivity of 23-26% was highly significant (p<0.001) for all three interns.
Conclusion:
Artificial neural networks can be trained to gain a performance in the interpretation of ST-segment changes in accordance to experienced cardiologists. The neural networks offers a reliably support to physicians with lesser experience in the interpretation of ECG with respect to changes that make the patient a candidate for reperfusion therapy.
To develop an automated tool for the analysis of electrocardiograms (ECG) with respect to changes that make the patient a candidate for reperfusion therapy. An additional aim was to assess the influence of the tool on the ECG classifications of three interns.
Methods and Results:
An artificial neural network was trained to interpret ECGs regarding changes making the patient a candidate for reperfusion therapy. The ECG measurements used as input to the network were obtained from the measurement program of the ECG recorders. The network was trained using a database of 3000 ECGs recorded at an emergency department. In the second step three interns classified 1000 test ECGs twice at different occasions, first without and thereafter with the advice of the neural network. The gold standard of the training and test ECGs was the classification of two experienced cardiologists. The three interns showed on average a sensitivity of 68% at a specificity of 92% without the advice of the neural network and a sensitivity of 93% at a specificity of 87% with the advice. The neural network itself showed a sensitivity of 95% at a specificity of 88%. The increase in sensitivity of 23-26% was highly significant (p<0.001) for all three interns.
Conclusion:
Artificial neural networks can be trained to gain a performance in the interpretation of ST-segment changes in accordance to experienced cardiologists. The neural networks offers a reliably support to physicians with lesser experience in the interpretation of ECG with respect to changes that make the patient a candidate for reperfusion therapy.
Avdelning/ar
Publiceringsår
2006
Språk
Engelska
Sidor
151-156
Publikation/Tidskrift/Serie
Clinical Physiology and Functional Imaging
Volym
26
Issue
3
Länkar
Dokumenttyp
Artikel i tidskrift
Förlag
John Wiley & Sons Inc.
Ämne
- Physiology
Status
Published
Forskningsgrupp
- Nuclear medicine, Malmö
ISBN/ISSN/Övrigt
- ISSN: 1475-0961