Application of Fractal Algorithms to Identify Cardiovascular Diseases in ECG Signals

摘要

The aim of this article was the identification of cardiovascular diseases, after applying Katz and Higuchi fractal algorithms on 4 databases of ECG signals downloaded from the Physionet website: heart failure (HF), hypertension (H), ischemic heart disease (IHD) and normal sinus rhythm (NSR). For this purpose, initially the ECG signals passed through a filtering stage using a Butterworth high pass filter of order 6 and 0.5Hz of cutoff frequency, in order to cancel variations of the baseline. The fractal algorithms were applied independently for each database. For such application, all signals were standardized with a total of 100,000 data, and for the calculation of each fractal dimension (FD) a frame equal to 10,000 with an overlap of 1,000 was used in a first stage; in a second stage, a frame equal to 1,000 with an overlap of 100 was used. Thus, the results showed that the Higuchi algorithm, in general, has a better performance compared with the Katz algorithm. These results refer to the observation of the variance of the FD averages, which are shown in Table 4. For example, in the cardiovascular disease Arterial Hypertension, the Higuchi algorithm presented 0.0093 of variance compared with the algorithm of Katz that only reached 0.0352. Complementarily, we used the Principal Component Analysis (PCA) to show graphically the differences between algorithms, so the components related to the Higuchi algorithm were presented in a less dispersed and grouped form, making possible their differentiation, mainly when the second analysis stage was carried out.