Correlation models for monitoring fetal growth

Area of Science:

• Maternal-fetal medicine
• Biostatistics
• Medical imaging

Background:

• Fetal growth monitoring is crucial for assessing pregnancy health.
• Standard charts are used to compare fetal growth against expected norms.
• Accurate assessment of fetal growth requires robust statistical models.

Purpose of the Study:

• To model the longitudinal dependence of key fetal biometry measurements.
• To develop a statistical method for evaluating fetal growth adequacy.
• To provide a tool for assessing fetal growth between 14 and 40 weeks gestation.

Main Methods:

• Utilized data from the Fetal Growth Longitudinal Study of the INTERGROWTH-21st project.
• Employed a two-stage modeling approach: data transformation to Z-scores and Gaussian process correlation modeling.
• Modeled longitudinal dependence for fetal head circumference, biparietal diameter, occipito-frontal diameter, abdominal circumference, and femur length.

Main Results:

• Developed a method to transform non-normal ultrasound measurements into standardized deviations (Z-scores).
• Fitted a Gaussian process correlation model to describe the relationship between measurements from 14 to 40 weeks.
• The resulting correlation structure allows for the assessment of growth between successive fetal measurements.

Conclusions:

• The developed statistical model provides a robust framework for analyzing fetal growth data.
• The correlation model enhances the ability to determine the normalcy of fetal development.
• A Shiny application is available to facilitate the practical application of these findings.