Linear time series analysis of long term fetal heart rate variability patterns

Area of Science:

• Biomedical Engineering
• Signal Processing
• Maternal-Fetal Medicine

Background:

• Long-term fetal heart rate variability (LFHRV) is crucial for assessing fetal well-being.
• Current methods rely on visual inspection of fetal heart rate (FHR) traces by healthcare professionals.
• LFHRV data is characterized as a non-stationary, random, and correlated time series.

Purpose of the Study:

• To develop a parsimonious stochastic model for the numerical representation of LFHRV.
• To objectively identify random cyclical patterns in detrended FHR data.
• To establish a quantitative method for analyzing LFHRV, complementing visual assessment.

Main Methods:

• Application of linear stochastic time series analysis techniques.
• Utilizing Maximum Likelihood estimation with Kalman filtering for parameter estimation.
• Analysis of short (2-minute) quasi-stationary contiguous blocks of LFHRV data.

Main Results:

• Identification of a second-order autoregressive (AR(2)) model as statistically adequate for LFHRV patterns.
• The AR(2) model effectively captures variability in 2-minute detrended average FHR data windows.
• Spectral analysis of the identified model allows for numerical detection of pseudo-periodicity in LFHRV.

Conclusions:

• A statistically adequate AR(2) model provides a quantitative representation of LFHRV.
• This numerical approach enables objective detection of random cyclical patterns and pseudo-periodicity in FHR.
• The developed method offers a data-driven alternative to visual inspection for fetal condition assessment.