An ECG feature extraction method based on GASF and MPRE2D for the detection of congestive heart failure

Congestive heart failure (CHF) is a cardiovascular disease that poses a serious threat to human health. Electrocardiogram (ECG) signals can be used to detect heart diseases such as CHF. However, the low amplitude and short duration of ECG signals severely affected CHF detection. This paper proposes a CHF detection method based on Gramian angular summation field (GASF) and two-dimensional multiscale permutation-ratio entropy (MPRE2D). First, ECG signals are preprocessed and converted into ECG images using the GASF algorithm. GASF can convert one-dimensional ECG signals into two-dimensional coded images containing important information. Then, the two-dimensional permutation-ratio entropy and MPRE2D algorithms are introduced to measure the irregularity and complexity of ECG images. Finally, the MPRE2D features of the image are extracted and the feature vectors are classified using a support vector machine. The classification accuracy is 99.46%, sensitivity 99.36%, specificity 99.63% and F1-score 99.56% on the normal sinus rhythm database and congestive heart failure database. Computer simulations show that the methods based on GASF and MPRE2D provide an effective method for CHF detection. This method can accurately detect patients with CHF using only 2 s of ECG signals length. It not only provides valuable references for clinical doctors to assess and treat CHF, but also offers clinically significant results for CHF risk assessment.