Adaptive Spectral Analysis of Umbilical Venous Doppler Maternal–Fetal Power Ratio for Preeclampsia Risk Assessment

This is the authors' abstract. We don't add key points for this paper.

Objectives To develop and evaluate adaptive spectral analysis for deriving the maternal–fetal power ratio (MFPR) from umbilical venous Doppler recordings and to assess MFPR for subsequent preeclampsia (PEC) risk discrimination across recording durations. Methods We retrospectively analyzed second‐trimester umbilical venous Doppler recordings in 2 datasets: 83 short acquisitions (mean, 3.4 seconds) from 46 subjects (15 with subsequent PEC), and 126 long acquisitions (mean, 6.9 seconds) from 66 subjects (30 with subsequent PEC). Power spectra were estimated by discrete Fourier transform, Welch, and Capon estimators with adaptive window selection, diagonal loading, and maternal and fetal heart‐rate‐guided peak localization. MFPR was defined as the decibel difference between the maternal and fetal spectral‐peak powers and averaged across clips for each subject (median, 2 clips per subject in both datasets). Discrimination was assessed using group differences, receiver operating characteristic area under the curve (AUC), and Youden's . Results Higher MFPR during the second trimester was associated with subsequent development of PEC. For short recordings, heart‐rate‐guided adaptive Capon performed best (AUC, 0.84; 95% confidence interval [CI], 0.70–0.95). For long recordings, heart‐rate‐guided Welch performed best (AUC, 0.75; 95% CI, 0.60–0.88). Adaptive window selection reduced performance sensitivity to fixed window length, and diagonal loading stabilized covariance inversion. Conclusions Umbilical venous Doppler MFPR distinguished pregnancies that later developed PEC. The best‐performing spectral‐analysis pipeline differed between the short and long acquisition datasets. Prospective validation is needed.