We hereby report results of preclinical validation of PhysECG reconstruction algorithm. The algorithm is based on four assumptions:
1) That the ECG in each lead: bipolar or unipolar is equivalent to spatial gradient of potentials - this means that the ECG is a measure of asymmetry of cardiac activity and not the activity itself.
2) The potentials entering this gradient resemble long known monophasic potentials that travel throughout the body towards all electrodes.
3) That by their physical nature travelling waves are longitudal waves, that represent displacement current, which changes density of surface charge in the sensitivity field of electrode and changes its potential.
4) That each monophasic potential can be represented as a simple superposition of cardiomyocyte potentials, which start at time instants, related to passage of activation wave under each electrode. Signal decomposition based on these assumptions and on a well-known cardiomyocyte model, such as Luo-Rudy or ten Tussher allows to distingush the contributions from individual electrodes: activation functions, which reflect the timing of passage of activation wave. It can be performed using AI-based algorithm. A draft of such algorithm was prepared in order to reconstruct ventricular activation functions, based on the QRS complex of the ECG. In our case the training set comprised of 19.000 recordings of PTB Diagnostic ECG Database and the control group consisted of 80 data entries. The preliminary results obtained using PhysECG lead to conclusion that we are able to observe left and right ventricle separately. We also identify gender changes. Furthermore, we can identify changes related to hypertrophy, conduction disturbances or ischemia, which require further validation, but seem to be consistent with the known facts on prolongation of muscle conductance. The results are stable across the study group. The algoritm is computationally simple, fast and stable, with particularly low demand for resources.