An algorithm to obtain the QRS score based on ECG parameters detection and neural networks for confounder classification

The present work proposes an algorithm to calculate the QRS Score and the determination of confounders starting from Electrocardiographic (ECG) signals. The QRS Score is a parameter that indicates how big the scar is in the wall of the patient’s myocardium; It is also helpful in determining how healthy the heart is. Said parameter is calculated from signal information such as time measurements, amplitude relationships and waveforms. The evaluation of the ECG signals is usually done by visual perception of the graph paper where it is printed as a result of the electrocardiogram examination. However, the reproducibility of this method is 60% and the repeatability is 66%. This definitely affects the accuracy of the score obtained and therefore the diagnosis of a disease. The proposed algorithm aims to reduce the subjectivity of the analysis and standardize the punctuations to be obtained. The algorithm is made up of processing stages that involve the conditioning of the signal using finite impulse response (FIR) filters, decontamination of confounders by neural networks, detection of the QRS complex, detection of times and amplitudes and finally obtaining the QRS score from a table of criteria. Finally, the proposed algorithm obtained a reproducibility of 75% and a repeatability of 100% exceeding the performance of the specialist.