Discriminant functions and decision tree induction techniques for antenatal fetal risk assessment

Abstract

This study concentrates on the comparison of the discriminant functions and the decision tree induction techniques in antepartum fetal evaluation. These classification techniques are applied to antenatal fetal risk assessment problem and the performances, the computational complexities and the importance of each technique in terms of diagnostic clues are observed. The task is to investigate the doppler ultrasound measurements of umbilical artery (UA) to relate the health conditions of fetuses using discriminant functions such as linear discriminant functions (LDF), multilayer perceptron (MLP), decision trees (C4.5, CART) and neural trees. We use the following UA blood flow velocity waveforms: pulsatility index (PI), resistance index (RI) and systolic/diastolic ratio (S/D) in terms of weeks (week index: WI as a normalized value) to decide if there is any hypoxia suspicion. It is observed that the performances of MLP and CART are better but C4.5 defines understandable diagnostic clues. On the other hand, the time complexity of LDF and C4.5 are becomes favorable. Experiments support that C4.5, MLP, CART and neural trees are favorable medical aids to phycisians during intensive surveilance of fetuses. With the limited number of indices, we obtain a specificity and sensitivity of 100% and 93% with these decision techniques.