ECG calibration signal database construction based on IEC 60601-2-25 using MATLAB

Abstract

ECG machines should be calibrated and tested to assure their accuracy. IEC 60601-2-25 standard described signals for calibrating ECG amplitudes and frequencies. The problem was this standard does not clearly describe the formulas of the calibration signals nor the complete database of these signals. The aim of this study was to get a database of ECG calibration signals for testing based on IEC 60601-2-25 standard clause 201.12.1.101. Data were constructed with a series of sinus function in Matlab software to simulate P, Q, R, S, and T segments. The data were visually and statistically compared with the data from commercial CTS database. Data were constructed for 3 different lead of 12 ECG calibration data. Four ECG calibration data with elevation or depletion ST segment were excluded from this study. This study demonstrated that these ECG calibration signals were slightly visually different and statically had differences in some of S wave and most of T wave. This data can be used by designers or manufactures, but for the testing laboratory is recommended using a commercial product.

A piecewise spline approach for modeling of ECG signals

“…Machine learning based models is also proposed in the ECG modeling such as BiLST, utilizes a generative adversarial network (GAN) composed of a bidirectional long short-term memory Moreover, ensuring the accuracy of ECG machines requires calibration and testing, a crucial aspect covered by the IEC 60601-2-25 standard. However, the standard lacks clear formulas for calibration signals and a comprehensive database of these signals and it is necessary to test the computerized ECG with reference ECG [24,25]. This research is also dedicated to addressing this issue by focusing on the fidelity of the modeled signal and obtaining a database of ECG calibration signals for testing, following the guidelines of the IEC 60601-2-25 standard clause 201.12.1.101.…”

“…Some ECG models based on generative adversarial network (GAN) have been also proposed which also require a large real database for training purposes [22,23]. The ECG models can be compared with IEC60601 standards as specified in [24,25]. Significant research work in context to different types of electrodes, response of amplifiers, coupling of the electrodes to the amplifier and the amplifier input impedance in view of IEC60601 standard has been described in [26,27].…”