Cardiotocographic monitoring was a common procedure used to assess the fetal status during pregnancy and labor. Although the monitoring and feature extraction of fetal heart rate signals could be done by automated computer systems, the final diagnosis of fetal state still largely depended on clinicians’ interpretation of the CTG features. In the current project, multinomial logistic regression models were used to model the relationship between CTG features and a three-level fetal outcome: normal, suspect, and pathologic. Proportional odds and baseline odds models were compared for their performance, and stepwise model selection procedure was conducted to identify proper predictors. The final models were able to provide decent goodness-of-fit and high classification accuracy.
yixlu/Cardiotocographic-Monitoring-and-Classification-of-Fetal-Outcome
Modeling the relationship between fetal heart rate signals and fetal outcome using multinomial logistic regression