New algorithms for continuous analysis of long term ECG recordings using symplectic geometry and fuzzy pattern recognition

In this paper, we describe how to analyze continuously ECG of any duration, and we describe a group of new and original algorithms which can do this analysis almost online. These algorithms use symplectic geometric calculus as well as an original fuzzy clustering method for localization of various waves and calculation of different parameters. The programs calculate the isoelectric curve of the ECG, then localize all ECG complexes and delineate the P, QRS and T waves within each complex. Then they calculate various parameters such as spatial area of the QRS wave, QRS-T angle etc. Finally, according to predefined criterions, every complex of ECG is classified into one of several predefined classes and an average prototype of every class is calculated and recorded. Spurious parts and outliers in continuous signal or in calculated parameters are automatically detected. The calculation of the isoelectric curve is performed by using an original fuzzy pattern recognition method operating on a detection function