Design and development of an annotated ECG database

Annotated databases are used in the evaluation and testing of the performance of electrocardiographic (ECG) processing and data compression algorithms. They help greatly in testing and evaluating new developed systems, as they require proper validation of their performance. The present work documents the design and development of an annotated, single-channel digital ECG database. The procedures for forming the annotated database consist of five processing steps. The first step is to eliminate the different types of noise interfered with the ECG signals using digital filters. The second step is the detection of QRS-complexes. The third step is to extract quantitative parameters identifying different ECG signals. The fourth step is the classification process in which the ECG beats are classified into different categories. The last step is the annotation process: each beat is given a specific symbol. Since the formation of the database itself has significant effect on the measured performance of the system under investigation, efficient techniques that give the highest performance must be selected for each of the mentioned processing steps. The developed database is similar to the known AHA database. It consists of fifty-five records, each of period of 10 minutes and is recorded on 1.44" floppy diskettes. The database contains five categories of ECG records: (1) ventricular arrhythmia, (2) atrial fibrillation/flutter, (3) ischaemic heart disease, (4) ECG test compression database, and (5) pacemakers. The developed database is a preliminary version for developing a complete Egyptian annotated database.