Instrumento virtual para implementação e testes de algoritmos de extração de parâmetros de ECG através da transformada wavelet

Abstract

The electrocardiogram is a medical procedure that has become very popular because it is accessible, non-invasive, easy to perform and, mainly, has high sensitivity for diagnosing heart diseases, including arrhythmias. A more detailed representation of a medical examination can facilitate the analysis of the health professional, making it easier to diagnose and treat illnesses in a higher quality of care for hospitals and better quality for patients. The Wavelet Transform is a very promising mathematical method for a time-frequency analysis that decomposes the signal into coefficients that describe its details at different instants of time, making a time-scale analysis. This work proposes the implementation of a system for testing algorithms and analyzing electrocardiographic signals in the form of a virtual instrument, to read a file in the standard CSV with those taken from a public database. The virtual instrument is implemented using LabVIEW software, which is responsible for reading, adjusting and processing the signal, detecting the QRS complex and T wave extraction parameters of the electrocardiographic signal using a Wavelet Transform for a time-scale analysis. Processing the Wavelet Transform and extracting the parameters is done by a MATLAB © script and developed in the virtual instrument. It also performs calculations to determine the patient's heart rate during the procedure and ST segment duration, which are important parameters for diagnosing heart disease. To validate the results, the algorithm analyzed samples of electrocardiographic signals taken from the ECGID Database, then the Root Mean Square Error was applied to the detected positions, which resulted in practically null values for the variation of the detected points in relation to the real positions of the QRS complex and a four-sample mean rate of change in the detection of the T wave relative to the real positions. In addition to the RMSE, another parameter applied to measure the quality of the results was the use of the sensitivity calculation, which resulted in 100% for the detection of QRS complex positions and 83% for the detection of the T wave, which could be justified by anomalies in the signal analyzed electrocardiographic and a small limitation of the algorithm regarding the detection of the last analyzed period.