Precise T-wave endpoint detection using polynomial fitting and natural geometric approach algorithm. (February 2023)

• Record Type: Journal Article
• Authors: Winkert, T.; Benchimol-Barbosa, P.R.; Nadal, J.
• Is Part Of: Biomedical signal processing and control. Volume 80(2023)Part 1
• Publication Date: 2023-02
• Subjects: T-wave -- T-wave endpoint -- QT interval -- Cartesian curvature -- Polynomial fitting -- Automated method; ECG Eletrocardiogram -- CGC Cartesian Geometric Curvature -- CGM Cartesian Geometric Curvature Method -- TTI Terminal T-wave Interval -- QTM Physionet QT Database Marks -- RMSD Root-Mean-Square difference -- RMS Root-Mean-Square; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.104254
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract: QT interval (QT) is defined as the distance between the beginning of the QRS complex and the end of the T-wave, and it reflects the time course of the ventricular depolarization and repolarization on the surface electrocardiogram (ECG). QT and its variability over time are modulated by the autonomic nervous system and are related to arrhythmogenesis. A challenging task for appropriate QT assessment is the detection of the T-wave endpoint. This study proposes a novel automatic approach to correctly identify the T-wave endpoint. After the peak of the T-wave, it was assumed that the quasi-asymptotical hyperbolic waveform decay of the T-wave acutely bends to meet the ECG baseline and then smooths out onto the TP segment. The point showing the maximal baseline bending is usually assumed by the cardiologists as the T-wave endpoint. In this approach, the terminal portion of the T-wave was represented by a parsimonious-optimal order polynomial function, in which the Cartesian curvature was calculated. One hundred and one ECG records from Physionet QT Database were analyzed to investigate the performance of the method. A Cartesian Curvature-based method (CGM) was developed and applied for automated detection of the T-waves endpoints. Q-waves were also measured automatically. The QTs were calculated and compared with respective cardiologist manual marks provided by QT Database by Pearson's correlation and Bland-Altman charts. High correlation (Pearson's Correlation = 0.94; p < 0.001) between the novel approach and the reference marks, observed in all analyses, showed the method's suitability to identify T-wave endpoints. The CGM performs as the cognitive experience of the cardiologist and has a simple mathematical implementation, indicating to be a promising tool for QT interval assessment.