Clustering with ϵ-Hyperballs Based Simplification of Fuzzy Rules to Support the Assessment of Fetal State

CardioTocoGraphy (CTG) is a fundamental procedure to support fetal diagnosis during pregnancy and labor, which involves registration and analysis of fetal heart rate and uterine contraction signals. As the fetal state assessment based on visual evaluation of the signal printouts is demanding, methods supporting the diagnostic decision are explored. To provide the qualitative assessment of the fetal condition, a fuzzy classifier can be applied. Its main advantages include the knowledge base (fuzzy rule base) linguistic interpretability. As fuzzy models involving fewer rules are easier to interpret, procedures of Rule Base Simplification (RBS) are the topic of research. This work describes the RBS method based on the fuzzy clustering with eps-hyperballs integrated with the evolutionary strategy framework. Its objective is to decrease a rule base size of zero order Takagi-Sugeno-Kang (TSK) fuzzy classifier while maintaining high CTG signals classification quality. The efficiency of the proposed RBS procedure was confirmed based on the evaluation of CTG signals from the benchmark CTU-UHB database.