Detection of Multilead ECG Character Points and Assessment Based on a Reference Database

摘要

A method of multilead ECG character points detection is introduced in this study. The method includes two main sections: the detection of single lead character points and determination of global character points. The method of single lead character points detection is similar to the traditional ones that use the first differential signal of ECG. While some special modification is adopted. The multilead ECG character points are gotten according the single lead result. To make the detection more reliable, an approach of lead optimization is used. In the process of global character points determination, such as the detection of the onset and the offset of QRS, P wave and T wave, the similar methods are used. Since some approaches can be found in several other papers, these processes are not explained into detail. The method is assessed with the CSE database. Since only part reference of CSE database is published, the test result is base on the MA1 data set which reference points are published. The result reveals that the method of detecting the fiducial point of P and QRS has similar accuracy as referee or average result of other computer program. While method of detecting the offset of T may need further improvement. In addition, to test the method stability and reliability, the character points of the CSE''s other data set (MA2, MO1, MO2) are also marked beat by beat manually. Then another assessment based on ourselves reference is also executed. The result is similar to the MA1 test result