Multivariate analysis based on linear and non-linear FHR parameters for the identification of IUGR fetuses

Area of Science:

• Obstetrics and Gynecology
• Fetal Medicine
• Biomedical Signal Processing

Background:

• Fetal Heart Rate (FHR) monitoring is crucial for detecting pathological fetal conditions during pregnancy.
• Intra Uterine Growth Restriction (IUGR) is a significant concern requiring accurate diagnostic tools.
• Existing FHR analysis methods may benefit from multivariate approaches for improved discrimination.

Purpose of the Study:

• To develop and validate a multivariate approach for discriminating between normal and IUGR fetuses using FHR signals.
• To identify a minimal set of FHR parameters with high discriminative power.

Main Methods:

• Collected FHR recordings from 120 fetuses (60 normal, 60 IUGR) using standard CTG non-stress tests.
• Computed 8 linear and non-linear FHR indices based on prior studies of their discriminative properties.
• Utilized the Orange® data mining suite to evaluate various multivariate discrimination models.

Main Results:

• A Logistic Regression model using only 4 selected FHR parameters achieved 92.5% accuracy.
• The model demonstrated high performance with 93% sensitivity and 91.5% specificity for identifying IUGR fetuses.

Conclusions:

• A multivariate approach based on a limited set of FHR parameters is effective for accurate fetal condition discrimination.
• This method offers a promising, non-invasive tool for identifying Intra Uterine Growth Restricted fetuses.