Automatic identification of movement and muscle artifacts in ECG based on statistical and nonlinear measures

This work aimed to develop an algorithm for the automatic identification of movement and muscle artifacts in ECG recordings based on a set of basic statistical features and the Higuchi fractal dimension (HFD). The use of a threshold for variance above 0.0225 and an HFD in the range from 1.11 to 1.26 allowed for the classification of one-second ECG segments with and without artifacts with sensitivity, specificity, precision, and negative predictive value of 66.3%, 99.8%, 75.2%, and 99.7, respectively. The accuracy of the Movesense device is very high (99.5%). The proposed method can be helpful in eliminating a significant number of artifacts in ECG signals monitored by Movesense.