Classification of fetal heart rate tracings based on wavelet-transform and self-organizing-map neural networks

The objective of the present study is the development of an automated computerized system that will assist the early diagnosis of fetal hypoxia. We demonstrate that it is possible to distinguish between healthy subjects and academic fetuses by way of wavelet transform analysis of the fetal heart rate recordings and fetal pulse oximetry (FSpO₂). We focus on the values of the standard deviation of the wavelet components (up to scale index 5) and we apply a Self-Organizing-Map in order to investigate the relationship between the fetal heart rate variability in different scales and FSpO₂ (taking as a threshold for the FSpO₂, the 30% level and considering the minimum value of FSpO₂ during a 10-minute segment) for normal and acidemic fetuses during the second stage of labor, which can be used to discriminate acidemic fetuses from normal ones. A total accuracy of 91% has been achieved, enabling us to correctly classify all the normal cases (but one) as belonging in the normal group and all pathologic cases (but two) as belonging in the acidemia group, therefore providing a clinically significant measure for the discrimination of the different groups. Fetal pulse oximetry seems to be an important additional source of information.