THE PREDICTION OF THE LOW FETAL BIRTH WEIGHT BASED ON QUANTITATIVE DESCRIPTION OF CARDIOTOCOGRAPHIC SIGNALS

Cardiotocography (CTG) is a routine method of fetal condition assessment used in modern obstetrics. It is a biophysical method based on simultaneous recording and analysis of activity of fetal heart, fetal movements and maternal uterine contractions. The fetal condition is diagnosed on the basis of printed CTG trace evaluation. The correct interpretation of CTG traces from a bedside monitor is very difficult even for experienced clinicians. Therefore, computerized fetal monitoring systems are used to yield the quantitative description of the signal. However, the effective methods, aiming to support the conclusion generation, are still being searched. One of the most important features defining the state of fetal outcome is the weight of the newborn. The presented work describes an application of the Artificial Neural Network Based on Logical Interpretation of fuzzy if-then Rules (ANBLIR) to evaluate the risk of the low birth weight using a set of parameters quantitatively describing the CTG traces. The obtained results confirm that the neuro-fuzzy based CTG classification methods are very efficient for the prediction of the fetal outcome. Cardiotocography (CTG) is a primary biophysical method for monitoring of fetal condition before and during labour. It consists in analysis of the fetal heart rate (FHR) variability and their relationship to fetal movements and maternal uterine contractions. At a present, CTG become a standard clinical technique for identifying a wellbeing of the fetus. However, visual analysis of graphical patterns describing the FHR variability is relatively difficult. Therefore, computerized fetal monitoring systems are used to provide the quantitative description of the CTG signal. Nevertheless, the interpretation is still made by clinicians and remains highly subjective and dependent on the human expert capability and experience. The repeatable and objective assessment of the fetal state is of particular importance. Early detection of fetuses that during pregnancy are at significant risk may help to avoid dangerous situations which are more difficult or even impossible to manage in case of the newborn. Consequantly, effective methods, aiming to support the diagnosis, are still the topic of many research studies [3], [8][10]. The CTG monitoring evaluates the actual (at the time of monitoring session) fetal state. But the diagnosis verification is possible only after the delivery. There is no reference information about the fetal health condition during pregnancy. This information will be obtained only after the delivery, and the fetal outcome is retrospectively assigned to the fetal state. Such prediction of fetal outcome during pregnancy is possible, because in perinatology it is assumed that the fetal state can not change rapidly. One of the most important features defining the fetal outcome is the newborn’s weight. Babies categorized as low birth weight are at particular risk of health problems, disability, or even death. In obstetrics, the newborns weight is evaluated as a percentile of reference values of birth mass in relation to the completed week of gestation. Any birth weight below the 10th percentile of the reference values is considered as too low . The presented work describes an application of the Artificial Neural Network Based on Logical Interpretation of fuzzy if-then Rules (ANBLIR) [4] to evaluate the risk of low birth weight using parameters for quantitative description of CTG traces. The ANBLIR is a computationally effective fuzzy reasoning system which connects advantages of neural networks (capability of learning and generalization) and fuzzy systems (capacity to handle inherently imprecise concepts and ability of linguistic interpretation of learning results). It was successfully applied to solve many practical problems leading to significant increase of performance in comparison to the solutions using classical computational intelligence algorithms [4], [3]. To establish values of the neuro-fuzzy system parameters for birth weight assessment we used a preprocessed database obtained from an archive of computerized fetal surveillance system MONAKO [7]. We investigated three different learning algorithms of ANBLIR based on integration of steepest descent method, least squares algorithm and deterministic annealing learning as well as different training data set structures in aim to achieved the best CTG classification accuracy.