Symbolic, Neural and Neuro-fuzzy Approaches to Pattern Recognition in Cardiotocograms

This paper describes several approaches to computer-supported recognition of accelerative and decelerative patterns in the Foetal Heart Rate signal with a view to automation of the diagnosis of foetal well being. The classifiers discussed evolve from a rule-based approach to a neuro-fuzzy system, via classical neural network architectures. The main problem regarding the symbolic approach is the limited possibility for knowledge elicitation. In order to resolve this problem and to obtain better limits for the decision regions, a classical neural network approach and a neuro-fuzzy system able to deal with ill-defined data were both investigated. The latter produced better results, with a classification rate of 72.6% for accelerations and 65.7% for decelerations. In the final section of our paper, we discuss how this approach opens up new lines of research, so as to improve the detection of accelerative and decelerative patterns.

[1]  Timothy Masters Signal and Image Processing with Neural Networks: A C++ Sourcebook , 1994 .

[2]  Sankar K. Pal,et al.  Neuro-Fuzzy Pattern Recognition , 1999 .

[3]  Vra Krkov Kolmogorov's Theorem Is Relevant , 1991, Neural Computation.

[4]  Jyh-Shing Roger Jang,et al.  ANFIS: adaptive-network-based fuzzy inference system , 1993, IEEE Trans. Syst. Man Cybern..

[5]  Amparo Alonso-Betanzos,et al.  Intelligent analysis and pattern recognition in cardiotocographic signals using a tightly coupled hybrid system , 2002, Artif. Intell..

[6]  Amparo Alonso-Betanzos,et al.  A Neural Network Approach to the Classification of Decelerative Cardiotocographic Patterns in the CAFE Project , 1998, NC.

[7]  Simon Haykin,et al.  Neural Networks: A Comprehensive Foundation , 1998 .

[8]  Bertha Guijarro-Berdiñas CAFE: un sistema con arquitectura híbrida para la monitorización inteligente del estado antenatal , 1999 .

[9]  Casimir A. Kulikowski,et al.  Computer Systems That Learn: Classification and Prediction Methods from Statistics, Neural Nets, Machine Learning and Expert Systems , 1990 .

[10]  Franco Scarselli,et al.  Are Multilayer Perceptrons Adequate for Pattern Recognition and Verification? , 1998, IEEE Trans. Pattern Anal. Mach. Intell..

[11]  Amparo Alonso-Betanzos,et al.  Adaptive pattern recognition in the analysis of cardiotocographic records , 2001, IEEE Trans. Neural Networks.

[12]  J. Parer Handbook of Fetal Heart Rate Monitoring , 1983 .

[13]  J.-S.R. Jang,et al.  Input selection for ANFIS learning , 1996, Proceedings of IEEE 5th International Fuzzy Systems.

[14]  Michio Sugeno,et al.  Fuzzy identification of systems and its applications to modeling and control , 1985, IEEE Transactions on Systems, Man, and Cybernetics.