On the correlation dimension of ECG signals

The analysis of ECG signals with methods derived from chaos theory is a potential tool to classify different heart behaviours and can help to get insights on the heart dynamics. Particularly, this work concerns the computation of the correlation dimension (D/sub 2/). It is well known that D/sub 2/ estimation requires some cautions when dealing with real signals affected by noise. In a previous paper the authors presented a method for a rough quantification of the reliability of the D/sub 2/ estimate through the definition of a parameter (/spl alpha/) easily obtained as a byproduct of D/sub 2/ computation. In this paper they use it to analyze several surface ECGs from MIT-BIH Arrhythmia Database. They observed a good agreement between an unreliable D/sub 2/ estimate (large /spl alpha/ value) and the noise level of the examined signals. This comparison is made possible by comments on signal quality available from the MIT-BIH directory.