Electrocardiographic (ECG) ischemic morphology with physiological plausibility may be synthesized using a newly developed technique published in the Journal of Electrocardiology.

In this study, the data from the STAFF III database which contains electrocardiographic (ECG) recordings from 59 patients undergoing an elective percutaneous transluminal coronary angioplasty (PTCA) procedure were analyzed. Recordings had a sampling rate of 1000 Hz and amplitude resolution of 0.6 mV. Data from leads II, III, and eVF were selected.

Data were preprocessed to minimize interference using low pass, high pass, and notch filters. ECG signals were assessed by an expert to select only patients who did not have other signs of heart disease. After these preprocessing steps, data from only 12 patients remained.

From these data, an ST-level time series was generated for evaluation of maximum and minimum ST-segment elevations. ST-level 1 (L1) was between 50 mV and 200 mV, L2, between 201 mV and 400 mV, L3 between 401 mV and 700 mV, and L4, >701 mV.


Continue Reading

A Gaussian model was used for parameter fitting for each ECG data section. With the selected parameters, a dynamic model based on 3 differential equations was used to reproduce ECG morphological changes observed during right coronary artery (RCA) occlusion. The developed model was assessed using real heartbeat parameters.

For beats with positive T-waves, the model had good mean correlation coefficients (cc) and positive predictive values (PPV) for: L1 at lead II (cc, 93.7%; mean, 125.3 mV; standard deviation [SD], 2.2 mV; PPV, 99.8%), lead III (cc, 78.9%; mean, 107.1 mV; SD, 6.4 mV; PPV, 98.7%), lead aVF (cc, 93.9%; mean, 92.8 mV; sd, 0.7 mV; PPV, 100%), and L2 at lead III (cc, 98.9%; mean, 268.8 mV; SD, 1.4 mV; PPV, 100%).

The model had better PPVs for beats with negative T-waves. ST levels L1, L2, and L3 had good predictions at all leads: lead II (PPV: 100% for all), lead III (PPV, 100% for all), and lead aVF (PPV, 99%; 100%; 100%, respectively). ST level L4 only had satisfactory predictive power at lead III (cc, 97.0%; mean, 776.4 mV; SD, 4.5 mV; PPV, 99.5%).

A limitation of this model is the extensive preprocessing and assessment by an expert cardiologist which resulted in a reduction of available data. For this reason, the investigators did not split the dataset into training and testing sets.

“[G]enerated ECG signals can be used to create an ischemic electrocardiogram database useful for several applications, such as delineation and segmentation, training and testing of automatic algorithms to detect and monitor cardiac ischemia, as well as novel filtering methods of the ECG signals,” concluded the study authors.

Reference

Soler A I R, Bonomini M P, Biscay C F, et al. Modelling of the electrocardiographic signal during an angioplasty procedure in the right coronary artery. J Electrocardiol. 2020;62:65-72. doi:10.1016/j.jelectrocard.2020.08.003