Intra-individual Variability of the Electrocardiogram: Assessment and exploitation in computerized ECG analysis

Computer interpretation of the electrocardiogram (ECG) was one of the first applications of computers in health care. The first systems were developed in the early sixties by Pipberger and Caceres. In the last decades, computerized ECG analysis has become one of the most widespread computer applications for decision support in health care. For example, over 50 million ECGs were analyzed by computer in the United States already in 1988. Large research efforts have been made to bring the performance of these systems to a level acceptable for routine use. In an international cooperative study, called 'Common Standards for Quantitative Electrocardiography' (CSE), most present-day ECG computer programs were evaluated. This evaluation was done by comparing measurements and interpretations computed by these systems with those of a panel of experienced cardiologists on the one hand, and with a diagnosis based on clinical evidence, not involving the EeG, on the other. It was concluded that ECG analysis programs can assist clinicians in achieving lllore unifornl and consistent interpretations of ECGs. However, continued testing and refinenlent was deelned necessary to enhance the perfonnance of these systenls. Despite their good diagnostic performance, ECG analysis programs suffer from a number of drawbacks that limit their practical utility. One of the most important shortcomings is the vulnerability for individual ECG variability. Identical ECG signals will result in identical measurements and interpretations, but small (and diagnostic inconsequential) differences between signals may result in an entirely different diagnostic interpretation. This variability can already be a problem when, e.g., multiple ECGs of the same patient are interpreted that have been recorded only minutes apart. Instable interpretations will not only considerably diminish the practical use of ECG computer prograrlls, users will also lose confidence in their performance. Furthernlore, variation in interpretations of shnilar ECGs suggests that the accuracy of programs can still be improved.

• View at Worldcat

Free Full Text ( Final Version , 3mb )