Session P72.6

Measurement Standards and Test Signals for Evaluation and Improvement of ECG Measurement Programs

S Hargittai*

Innomed Medical Inc.
Budapest, Hungary

Introduction: The most important step of quantitative ECG analysis is to determine the wave boundaries. It is very difficult to find objective evaluation methods without commonly accepted strict definitions for the ECG wave boundaries. For this reason, the CSE Working Party and the EN 60601-2-51 recommend using reference databases and test signals. The main objectives of this paper are to consider the limitations of using CSE databases and standard test signals for evaluation and improvement of ECG wave recognition programs.
Method: Three databases were used for evaluating the ECG measurement programs. The three-channel CSE library consists of real curves with independent reference points obtained by the referees for all leads. The multilead CSE library also consists of real curves, but has only one common onset and offset per record for all the leads. The third set contains the artificial analytical signals of the standard. These databases were used not only for automatic assessment, all records and channels were checked individually and deviations from the references were analyzed.
Results: Assessment of one-channel accuracy can be performed only by using the three-channel library and the standard analytical curves. Unfortunately, even the references of three-channel database cannot be applied without limitations. Since there is no commonly accepted strict medical and mathematical definition of the wave onset and offset, the referees put the reference points on base of their own visual judgment. Furthermore, nonstandard gain and time scales had to be used. Checking the reference onsets and offsets of the three-channel library did not reveal the strict mathematical rules that have been used by referees reviewing the curves. Moreover there are several arguable and inconsistent decisions. The multilead database cannot be considered validated because most of the reference points were determined not by referees. For the standard analytical curves the problem is with the continuity of the high-order time derivatives.
Conclusion: The CSE databases and the standard analytical ECG curves can be used for automatic evaluation of an ECG measurement program, but reference data should be used with care. Design engineers have to have appropriate knowledge in both the signal-processing field and in electrocardiology.

(Abstract Control Number: 72)