Analysis of fetal heart rate variability and stress conditions in fetuses using wavelets and neural networks (ALOPEX): A feasibility study

Fetal electrocardiography in pregnancy and labor has been considered a further step when compared with older subjective methods such as stethoscopic examination and visual observation of fetal movements in understanding the condition of a fetus in the various stages of pregnancy and labor. This becomes vital especially during the last hour of labor. The gynecologists need to diagnose very precisely whether the baby is acidemic, the type of acidemia and the severity of the fetal distress which is caused by this condition and make a decision to perform a cesarean section to save the baby's life. However, the fetal electrocardiogram (ECG) does not seem precise enough to be able to provide this kind of diagnosis by itself and sometimes it causes false alarms and unnecessary cesarean sections. In this dissertation, we have proved that, 77% of the acidemic fetuses could be detected successfully during the last hour of labor by analyzing the fetal heart rates with the wavelet transform and neural networks trained by (ALOPEX). Root mean square energy values of the wavelet coefficients are the inputs to the neural network. In addition to this, we have analyzed the same database with the combination of wavelet moments and ALOPEX in order to make a comparison between the performances of the RMSEs and the moments of the same wavelet coefficients and ALOPEX in detecting the acidemic fetuses. We have succeeded in detecting 77% of the abnormal fetuses with the combination of wavelet moments and ALOPEX while the rate of the diagnosis of the abnormals was only 60% with the wavelet RMSEs and ALOPEX.; Since we have diagnosed 77% of the acidemic fetuses correctly, we believe that the combination of wavelet moments and ALOPEX is a reliable method of analysis in identifying the acidemia-based stress conditions in fetuses in eliminating unnecessary cesarean sections.