In this paper, a novel real-time algorithm for detecting ischemia in the ECG signal is proposed. The goal of this research is to meet the requirements of some smart cardiac home care devices, which can automatically diagnose the ECG and detect the heart risks outside the hospital, especially heart ischemia without symptoms in their early stages. The algorithm is developed based on a real time R peak detector, time domain traditional ECG parameters, the advanced morphologic parameters from Karhunen-Loève transform, and the adaptive neurofuzzy logic classification. Besides, in order to improve the reliability of our algorithm, several significant constraints of the ECG signal are considered. As a result, the ischemia episodes can be detected if the ischemic alteration persists longer than one minute in the ECG signal.

In following the directive "ambulant medical treatment is preferable to in hospital treatment" the home care area is becoming more and more important in western societies. In this paper a system for monitoring cardiological risk patients is presented. Special attention was paid to an automatic detection of life-threatening events and the initiation of immediate help.

The algorithms used in automated external defibrillators (AED) do not provide the necessary reliability of detection of fibrillation, especially in presence of artifacts. The system described in this article allows detecting different artifacts, improving the detection quality. Also an approach is suggested, to extend the AED functions to detect pulseless electrical activity. The new algorithm uses the impedance signal as an additional parameter to diagnose the cardiac arrest. The parallel analysis of ECG and impedance signals allows detecting noise, motion, respiration and hemodynamic parameters. The comparative tests of the algorithm demonstrated the excellent performance of the algorithm.

Abstract

Sl'MMAUY In numy hranchcs of medicine it is iK\vssarv l·) carry out anulysis of ECG signal aiitnmatk'allv. It is particularly important in such jpplic.it ions äs emergency medicine, home care inaikmc and paticnt monitoring. In thi.s sludy an algorithm was clevcloped for the vlck\tion öl shockahlc urrhythmias to be uscd in antonuitk oxtcrnal detlbrillators. Vulidution öl the algorithm provcs that it h äs better pcrlonnantv, than the existing devices. SU iNAL PRE-PROCESSINGIn ihc lirst step the ECG is filtered with a notch filter to eliimnate the 50-Hz noise; the second stage is a band pass tilter with a frequency ränge of 1-30 Hz. At the same time signal polarity is checked (which is necessary lo detcet the wrong position of electrodes). All lilters are real i/cd on the base of bit shift and addition operations, so that this signal pre-processing gocs last and special DSP functions are not necessary. DOMAINOne of the main parts of ECG analysis is QRS-complex deteetion. The QRS-detector used is based on the algorithm from Tompkins [4,7]. To improve deteetion of the R-peak, special signal processing is used. The square of derivative of the filtered ECG signal passes through the MWI (moving window integrator). The windovv width of the filter correlates to the heart rate, improving the deteetion quality. MWI Output signal goes to the local peak detector, which provides "l" on the Output if the sign of the differcnce of two consccutive samples differs from the previous value. The QRS-detector itself gets the filtered ECG signal and a binary Output of the local peak detector. It is only activated, if a peak in the ECG signal is detected. If no peak is detected for a certain time, then the search-back algorithm is activated. So the peaks with lower ampütude can also be detected. The next step of the algorithm removes double beats. Double beats can be detected, for example, if the amplitude of T-wave is too high. Beat events are generated äs the Output of QRS-detector. The heart rate is also evaluated in this block. If the QRS-detector determines, that the input ECG signal is not stable (for example, R-R interval is irregulär, R-peak amplitude is too low, etc.), the "Low signal quality" flag is activated. This flag is of great importance for the decider block of the algorithm. Another important parameter, that correlates to the ventricular fibrillation, is "Percentage of time above threshold", or PTABT. PTABT parameter is a characteristic of ECG signal morphology; thus, the normal ECG has very small PTABT, and it is definitely greater for fibrillation. The filtered ECG signal is also used to detect asystole. If the ECG amplitude is less than 200 for a certain period of time, the signal is qualified äs asystole and no shock is recommended. FREQUENCY DOMAINTo determine the parameters described above, time domain signal processing is used. The tests show, however, that it is possible to increase deteetion quality (sensitivity and specificity) by adding frequency domain signal processing. To calculate the signal spectrum, mul...