Fuzzy System for Retrospective Evaluation of the Fetal State

Cardiotocography is the most common method of biophysical assessment of fetal condition based on the analysis of fetal heart rate (FHR) signal. Due to difficulties with automatic interpretation of recordings, artificial intelligence methods are frequently used for FHR signal classification. However, the problem is the evaluation of the true actual fetal state, that could serve as reference in learning algorithms. The prognostic value of the recorded signal can be objectively verified on the basis of retrospective assessment of the neonatal outcome, which is determined with a help of newborn attributes. In practical applications, only one selected attribute is usually used as the reference. Consequently, the information of the true actual neonatal outcome represented by the remaining attributes is lost. The paper presents a fuzzy method of the neonatal outcome evaluation as a function of all available newborn attributes. The consistency of inference results with the assessment based on single newborn attributes shows the higher effectiveness of the fuzzy system and indicates the possibility of practical application to the objective validation of the learning algorithms.